loT and

Machine Learning
for Smart Applications

Edited by G. Vennira Selvi, T. Ganesh Kumar,

AA. Prasad, Raju Hajare, and Priti Rishi
 IoT and Machine Learning for
Smart Applications

This book provides an illustration of the various methods and structures that are
utilized in machine learning to make use of data that is generated by IoT devices.
Numerous industries utilize machine learning, specifically machine learning-as-a-
service (MLaaS), to realize IoT to its full potential. On the application of machine
learning to smart IoT applications, it becomes easier to observe, methodically ana­
lyze, and process a large amount of data to be used in various fields.
Features:

• Explains the current methods and algorithms used in machine learning and
IoT knowledge discovery for smart applications
• Covers machine-learning approaches that address the difficulties posed by
IoT-generated data for smart applications
• Describes how various methods are used to extract higher-level information
from IoT-generated data
• Presents the latest technologies and research findings on IoT for smart
applications
• Focuses on how machine learning algorithms are used in various real-world
smart applications and engineering problems

It is a ready reference for researchers and practitioners in the field of information

technology who are interested in the IoT and Machine Learning fields.
IoT and Machine Learning
for Smart Applications

Edited by
G. Vennira Selvi, T. Ganesh Kumar, M. Prasad,
Raju Hajare, and Priti Rishi

CRC Press
Taylor &. Francis Group
Boca Raton London New York

CRC Press is an imprint of the

Taylor & Francis Group, an informa business

A CHAPMAN & HALL BOOK

Designed cover image: ShutterStock

First edition published 2025

by CRC Press
2385 NW Executive Center Drive, Suite 320, Boca Raton FL 33431

and by CRC Press

4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

CRC Press is an imprint of Taylor & Francis Group, LLC

© 2025 selection and editorial matter, G. Vennira Selvi, T. Ganesh Kumar, M. Prasad, Raju Hajare, and
Priti Rishi; individual chapters, the contributors

Reasonable efforts have been made to publish reliable data and information, but the author and publisher
cannot assume responsibility for the validity of all materials or the consequences of their use. The authors
and publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any future
reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter
invented, including photocopying, microfilming, and recording, or in any information storage or retrieval
system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com or
contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750­
8400. For works that are not available on CCC please contact [email protected]

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used
only for identification and explanation without intent to infringe.

ISBN: 9781032621081 (hbk)

ISBN: 9781032623269 (pbk)
ISBN: 9781032623276 (ebk)

DOI: 10.1201/9781032623276

Typeset in Times
by Newgen Publishing UK
Contents
About the Editors .......................................................................................................... vii
Contributors .................................................................................................................... ix

Chapter 1 Recent Advances in Machine Learning Strategies and

Its Applications................................................................... 1
S. Saranyadevi and Lekshmi Gangadhar

Chapter 2 Understanding the Concept of IoT ....................................................... 19

C. Emilin Shyni, Megha Menon K., and
Anesh D. Sundar ArchVictor

Chapter 3 Unlocking the Power of IoT: An In-Depth Exploration ..................... 32

Prakash Shanmurthy, Damodharan D.,
Anandhan Karunanithi, Pajany M., and J. John Bennet

Chapter 4 Machine Learning in Internet of Things............................................... 67

Sujni Paul and Grasha Jacob

Chapter 5 Role of Machine Learning in Real-Life Environment 81

Balasaraswathi V.R., Nathezhtha T., Vaissnavie V.,
and Rajashree

Chapter 6 Efficient Blockchain-Based Edge Computing System

Using Transfer Learning Model ............................................................97
P. Sivakumar, S. Nagendra Prabhu, S.K. Somasundaram,
J. Uma Maheswari, and Murali Murugan

Chapter 7 Introducing a Compact and High-Speed Machine

Learning Accelerator for IoT-Enabled Health
Monitoring Systems ............................................................................. 110
Kavitha V.P., Magesh V., Theivanathan G., and
S.S. Saravana Kumar

v
 vi Contents

Chapter 8 Realization of Smart City Based on IoT and AI................................ 127

□. SoftArchive

Chapter 9 Sentiment Analysis of Airline Tweets Using Machine Learning

Algorithms and Regular Expression............................ 143
S. Nagendra Prabhu, A.P. Rohith, Shubhankar Bhope,
and P. Sivakumar

Chapter 10 Smart Workspace Automation: Harnessing IoT and AI for

Sustainable Urban Development and Improved Quality of Life..... 163
Sapna R., Preethi, Pavithra N., Manasa C.M., and
Raghavendra M. Devadas

Chapter 11 Application of Digital Image Watermarking in the

Internet of Things and Machine Learning ................... 178
K. Prabha and I. Shatheesh Sam

Index ............................................................................................................................ 197

About the Editors
G. Vennira Selvi is Professor at the School of Computer Science and Engineering
and Information Science at Presidency University, Bengaluru, India. She received her
Doctorate Degree in Computer Science and Engineering from Pondicherry University,
Pondicherry in 2018 and M.E. degree in Computer Science and Engineering from
Manonmaniam Sundaranar University, Tirunelveli. She has published 20 papers in
reputed international journals, 6 papers as part of international conferences, and 12
papers as part of national conferences. She has published two books and nine book
chapters. She has received eight Indian patents, one Design grant patent, and one
international patent. She has 22 years of teaching experience in undergraduate and
postgraduate courses. Her current research areas are IoT, wireless sensor networks,
machine learning, and artificial intelligence.

T. Ganesh Kumar works as Associate Professor in the School of Computing Science

and Engineering at Galgotias University, NCR, Delhi. He received an M.E. degree
in Computer Science and Engineering from Manonmaniam Sundaranar University,
Tamil Nadu, India. He completed his full-time Ph.D. degree in Computer Science
and Engineering at Manonmaniam Sundaranar University. He was Co-Investigator
for two Government of India-sponsored funded projects. He has publications in many
reputed international Science Citation Index and Scopus indexed journals and confer­
ence proceedings. He has published more than ten Indian patents.

M. Prasad received his B.Tech, M.Tech, and doctorate degrees in Information

Technology, Information Security, and Mobile Communication from Pondicherry
University, Pondicherry, in 2008, 2010, and 2017, respectively. He is currently
working in VIT Chennai campus School of Computing Science and Engineering.
His current areas of research include IoT, security in mobile communication, cyber
physical systems, machine learning, and artificial intelligence. He is having various
professional memberships like IEEE, ACM, CSI, and Internet Society of India. He
has published in more than 20 journals and holds 10 patents to his name.

Raju Hajare is Professor at the BMS Institute of Technology and Management. He is

a doctorate in the field of “Nano Devices Modeling and Simulation”. He has 19 years
of academic and 2 years of industry experience at various levels of organizations.
Dr. Raju has published more than 30 quality research publications in reputed Scopus-
indexed international journals and IEEE proceedings. Professor Raju has two
textbooks to his credit in the domain of electronics.

vii
viii About the Editors

Priti Rishi is Associate Professor at the College of Engineering and Technology,

SRM Institute of Engineering and Technology, Vadapalani Campus, Chennai. She
received an M.E. degree in Electronics and Communication Engineering from Thapar
Institute of Engineering and Technology, Punjab, India. She completed her part-time
Ph.D. degree at the Faculty of Information and Communication Engineering at Anna
University Chennai. She has multiple publications to her credit and holds more than
three Indian patents.
Contributors
Anesh D. Sundar ArchVictor Theivanathan G.
Hodos Institute Velammal Engineering College
Kirkland, Washington, USA Tamil Nadu, India

J. John Bennet Lekshmi Gangadhar

School of Computer Science and Nanodot Research Pvt. Ltd.
Engineering and Information Trivandrum, India
Science
Presidency University Grasha Jacob
Bengaluru, India Government Arts and Science College
Tamil Nadu, India
Shubhankar Bhope
Department of Computational Megha Menon K.
Intelligence Department of Computer Science and
School of Computing Engineering Reva University
SRM Institute of Science and Bengaluru, India
Technology
Chennai, India Anandhan Karunanithi
Department of Computer Science and
Manasa C.M. Engineering
Department of Computer Science & GITAM School of Technology
Engineering GITAM University
Presidency University Bengaluru, India
Bengaluru, India
S.S. Saravana Kumar
Damodharan D. Srinivasa Institute of Technology &
Department of Computer Science and Science
Engineering Jawaharlal Nehru Technological
Galgotias University University Kakinada
Greater Noida, India Andhra Pradesh, India

Raghavendra M. Devadas Pajany M.

Department of Information School of Computer Science and
Technology Engineering and Information Science
Manipal Institute of Technology Presidency University
Bengaluru, India Bengaluru, India
Manipal Academy of Higher Education
Manipal, India

ix
x Contributors

J. Uma Maheswari Sapna R.

School of Computer Science and Department of Information Technology
Engineering Manipal Institute of Technology
Vellore Institute of Technology Bengaluru, India
Chennai, India Manipal Academy of Higher Education
Manipal, India
Murali Murugan
Macy’s Inc. Rajashree
Georgia, USA Vellore Institute of Technology
Chennai, India
Pavithra N.
Department of Computer Science & A.P. Rohith
Engineering Department of Computational
Manipal Institute of Technology Intelligence
Bengaluru, India School of Computing
Manipal Academy of Higher Education SRM Institute of Science and
Manipal, India Technology
Chennai, India
Sujni Paul
CIS Division I. Shatheesh Sam
Dubai Men’s Campus Department of PG Computer Science
Higher Colleges of Technology Nesamony Memorial Christian College
Dubai Academic City, UAE Tamil Nadu, India

K. Prabha S. Saranyadevi
Department of Computer Science Department of Biotechnology
Arunachala Arts and Science (Women) Paavai Engineering College
College (Autonomous)
Tamil Nadu, India Namakkal, India

S. Nagendra Prabhu Prakash Shanmurthy

Department of Computational School of Computer Science and
Intelligence Engineering and Information Science
School of Computing Presidency University
SRM Institute of Science and Bengaluru, India
Technology
Chennai, India C. Emilin Shyni
School of Computer Science and
Preethi Engineering and Information
Department of Information Technology SciencePresidency University
Manipal Institute of Technology Bengaluru, India
Bengaluru, India
Manipal Academy of Higher Education
Manipal, India
Contributors xi

P. Sivakumar Vaissnavie V.
School of Computer Science and Sathyabama Institute of Science and
Engineering Technology
Vellore Institute of Technology Chennai, India
Chennai, India
Kavitha V.P.
S.K. Somasundaram SRM Institute of Science and
School of Computer Science and Technology
Engineering Vadapalani Campus
Vellore Institute of Technology Chennai, India
Vellore, India
Balasaraswathi V.R.
Nathezhtha T. Vellore Institute of Technology
Vellore Institute of Technology Chennai, India
Chennai, India
 Recent Advances
1 in Machine Learning
Strategies and Its
Applications
S. Saranyadevi and Lekshmi Gangadhar

1.1 INTRODUCTION
The domain of computer science called machine learning (ML) aids machines in
learning without explicitly predetermined [1]. Rather than utilizing programing to
perform tasks, ML comprises the conception of algorithms. Machines pick up know­
ledge from historical patterns and examples from their previous experience. Models
are used to estimate new values can be constructed. When the amount of data and
queries is too great for natural solutions, ML can be used to help identify answers by
analyzing the data. It can make it easier for users to find crucial information faster.
Because machines learn faster and can even outclass humans in a few sectors, they
can easily tackle complex issues [2]. Its demand is constantly rising as a result. ML is
growing in relevance along with big data and cloud computing because of how many
problems it can solve with its computational capacity. It has a wider variety of uses
in other fields, particularly because it mainly helps in the unearthing of novel drugs
and enables physicians to diagnose patients accurately, enabling the early detection
of numerous ailments.
In the field of data analysis and computing, artificial intelligence (AI), particularly
ML, can be expanded quickly nowadays, usually enabling the applications to perform
intelligently [3]. In general, ML is the well-liked new technology in the fourth indus­
trial revolution (Industry 4.0) since it makes systems capable of automatically learning
from experience and improving without the need for special programing [4-6].

1.2 MACHINE LEARNING TYPES

There are various categories of learning based on a range of potential circumstances
regarding the accessibility of exams, training, and instructional materials, as outlined
in the following sections.

DOI: 10.1201/9781032623276-1 1
2 IoT and Machine Learning for Smart Applications

FIGURE 1.1 Supervised learning.

1.2.1 Supervised Learning

One of the most extensively utilized learning techniques is supervised learning (SL),
which involves feeding a data set as input and knowing the outputs for each matching
input. As seen in Figure 1.1, the ML model utilizes them to attempt to develop a rela­
tionship between the feed and the result. There are two types of SL methods: “clas­
sification” and “regression” problems. The ML model translates the incessant output
function toward the input variables in a regression issue. For example, the model
needs to estimate a person’s age based on an image of the person. The ML model
attempts to map input variables to discrete types when it comes to classification
problems. For instance, estimating home values according to location [7].

1.2.2 Unsupervised Learning

We have an edge over all other learning algorithms with unsupervised learning (USL)
since it attempts to solve problems whose outcomes and the influence of factors are
unknown to us. As exemplified in Figure 1.2, the structures are found by grouping
the provided data according to the correlation between the data’s variables. In this
case, the ML model’s primary job is to group unsorted data according to similarities,
patterns, and variations without requiring prior knowledge of the training set. The
machine is restricted in its ability to identify the veiled structure in unlabeled data on
its own. USL issues are categorized into: Clustering and association [7].
Clustering: In a clustering issue, we arrange the data points based on the
connections between the variables. For example, we can group customers based on
how they buy things from the stores. This approach is commonly employed for stat­
istical examination of data across several domains. Association: In an association
Machine Learning: Advances and Applications 3

FIGURE 1.2 Unsupervised learning.

problem, we find patterns that characterize a significant section of the dataset, like
predicting that buyers of X will probably purchase Y.

1.2.3 Semi-supervised Learning

An algorithm for semi-supervised learning (SSL) combines SL and USL. In ML
sectors where they have unlabeled data and getting the labeled data from this is a
laborious procedure, but it can be quite helpful [7]. We have a massive amount of
input data in SSL Problems; a small portion is labeled, and the remainder is left
unaltered. For instance, consider a photo library where only a few pictures (such as
those of an elephant, automobile, or bridge) have labels; the remaining pictures are
all unlabeled (Figure 1.3).

1.2.4 Reinforcement Learning

Reinforcement learning (RL) acts as a technique that produces an increasing number
of positive outcomes. The response that the atmosphere delivers to the learning algo­
rithm after it selects an output for a given input tells the algorithm how well the output
lights the learner’s objectives. Trial and error search and deferred outcomes are the
two major components that regulate whether RL occurring [8]. Another major com­
ponent in reinforcement learning (RL) is the reward signal (R). The reward signal
provides feedback on the agent’s actions in terms of how favorable or unfavorable
they are in achieving the desired outcome. It serves as a measure of success or failure,
4 IoT and Machine Learning for Smart Applications

FIGURE 1.3 Strategies of machine learning.

encouraging behaviors that yield higher rewards while discouraging those leading to
lower or negative rewards. This component is crucial in shaping the agent’s behavior
over time, as it learns to optimize actions that maximize cumulative rewards in the
long run.

1.2.5 Multitask Learning

The primary aim of multitask learning is to assist other learners to help them advance
and accomplish better. When this method is used on a given job, it essentially retains
the memory of the steps involved in solving that specific problem as well as the way
the learning algorithm responds to it to draw that conclusion. Similar techniques are
employed by the algorithm to solve other comparable tasks. By allowing all students
to share their experiences, the learning algorithm can be improved and executed more
successfully, allowing for simultaneous and efficient learning [9].

1.2.6 Ensemble Learning

In ensemble learning (EL), many separate learners are employed to create a single
learner. This separate learner could be a decision tree, neural network, or Naive
Bayes. Since the 1990s, ensemble learning has gained popularity. When it comes to
completing a particular activity, a group of learners is always preferable to an indi­
vidual [10].

1.2.7 Neural Network Learning

The biological reality of neurons, which are structures inside our brains that
resemble cells, is where essentially the neural network starts. Understanding how
neurons work is necessary for understanding brain networks. A neuron comprises
four parts: The dendrites, the nucleus, the soma, and the axon. The electrical
signals are picked up by the dendrites and transmitted to soma for processing.
Machine Learning: Advances and Applications 5

This mechanism results in a drift to the dendritic. Terminals that guide the output
to the subsequently linked neuron with the aid of an axon. The neural network that
allows electrical signals to flow throughout the brain comprises these connections
between neurons. An artificial neural network (ANN) functions similarly and has
three layers [11]. Three primary forms of ANNs exist: Reinforcement, unsuper­
vised, and supervised [12].

1.2.8 Instance-Based Learning

With this kind of learning, the student is sufficiently equipped to be capable of picking
up patterns that attempt to apply to the material that it is fed. Thus, it goes by this
name. The same lazy learner waits for the test set to reach before processing the
training dataset with it. The drawback is that as data volume increases, so does the
complexity of the system.

1.3 STATE OF THE ART

1.3.1 Real-World Problems

Computer-based systems can leverage all client information via ML. It responds to
commands from the program and adjusts to novel circumstances or modifications.
Algorithms display previously unprogrammed characteristics and adjust to the
data. A digital associate can scan emails and extract important ideas by learning to
read and recognize context. This kind of learning involves having the ability to pre­
dict future consumer behavior. It enables you to be proactive as opposed to reactive
and to have a deeper insight into your customers. ML can grow over time and applies
to many different sectors and industries. The real-world uses of ML are exemplified
in Figure 1.4.

1.3.2 Cybersecurity
Internet is prevalently employed for information and services alike. According to the
author [13], nearly 48% of the world’s populace has utilized the Internet as a source
of acquaintance since 2017. In developed nations, this percentage rises to 82%, as the
study [13] concludes. The Internet is a connectivity of various devices, networks, and
computers. Its primary function is to transfer data between devices via networks. The
improvements and breakthroughs in computer systems and mobile device networks
caused a surge in Internet usage. Since most people use the Internet to obtain data,
cybercriminals are more likely to target it [13]. It is considered stable when a
computer-based system delivers idea integrity, obtainability, and secrecy. According
to the study [14], if someone gains illegal access to the network to interfere with
consistent operations, the integrity and security of the computer-based system would
be jeopardized. Cybersecurity (CS), user assets, and cyberspace can be dwindling
from unauthorized discrete outbreaks and access. According to a study by Thomas
et al. [15], maintaining data availability, integrity, and confidentiality is the major
objective of CS.
6 IoT and Machine Learning for Smart Applications

FIGURE 1.4 Real-time uses of machine learning.

1.3.3 Healthcare
Many changes are occurring in the domains of industry, transportation, and gov­
ernment because of developments in the area of deep learning (DL) and ML. Over
the decades, many studies have been conducted on DL. DL has been used exten­
sively and has produced state-of-the-art results in several areas, like computer vision,
text analytics, and speech processing [16]. Recently, DL/ML strategies have been
executed in the healthcare field, and researchers have shown remarkable results in
tasks, such as brain tumor division, medical image reconstruction, lung nodule iden­
tification, types of lung disease, the identity of body parts, etc. [17].
CAD systems might deliver a second view that will help radiologists in corrob­
orating the diagnosis. DL and ML will progress the functionality of CAD and other
systems that will help radiologists in making judgments. Technological advancements
in big data, cloud-computing, edge-computing, and mobile connectivity also facili­
tate the usage of ML and DL algorithms in the field of healthcare [18]. Together,
they can increase prediction accuracy and promote the development of an intelligent,
human-centered solution.

1.3.4 Intelligent Transportation System

Intelligent transportation is an outcome of implementing communication, infor­
mation, and sensing technology in transit and transportation systems. Intelligent
Machine Learning: Advances and Applications 7

transportation systems, which include driverless cars, public transportation system­

based management, and traveler information systems followed by road traffic man­
agement, are essential components of smart cities. These services are anticipated to
have a significant positive impact on society by reducing contamination, increasing
energy effectiveness, and improving transit and transportation efficacy, which ultim­
ately improves traffic and road safety. Using techniques like RL and DL, which have
gained popularity recently in DL and ML models, allows for the correct generation of
decisions and predictions by utilizing patterns [19].

1.3.5 Renewable Energy

A growing number of people are looking for eco-friendly and other forms of energy
due to the detrimental impacts of burning fossil fuels and the fast depletion of these
resources. As stated in this chapter, the energy industry is expanding its usage of
renewable energy sources, including biomass, wind, solar thermal, tidal waves, geo­
thermal, and solar photovoltaic. There will be variability in the electrical networks
for several reasons, such as when supply exceeds demand or when demand exceeds
supply. ML is applied to energy management and optimization [20]. Table 1.1 exem­
plifies the different kinds of ML strategies with examples.

1.3.6 Smart Manufacturing and Grid

Various types are developed for manufacturing, and it is termed “Smart Manufacturing”,
which uses digital technology, worker training, fast design modifications, and high
adaptability in its manufacturing process. Efficient production recycling, optimization

TABLE 1.1
Different ML Strategies with Examples

Types of
S. No. Examples Model Development Learning Strategies
1 Regression and Models that utilize labeled Supervised Task-driven
Classification data for learning
2 Reduction of Models that utilize unlabeled Unsupervised Data-driven
dimensionality, data for learning
association,
and clustering
3 Category and By utilizing joint data Semi Labeled and
clustering models were developed supervised unlabeled
by employing the concept
of penalty and reward
as a base for model
development
4 Category and Reinforcement Environment-
control driven
8 IoT and Machine Learning for Smart Applications

of supply chains, and demand-driven, fast adjustments to production levels are among
the other duties. Smart manufacturing is made possible by advancements in robotics,
massively parallel processing power, and industry-wide networking for gadgets and
services. Despite criticism of the electrical energy grid’s basic architecture, this
design has endured consistently over time. The electrical system remains unchanged
over the centuries, and we are living in the 21st century. But power has become more
and more necessary as population and consumption have increased [21].

1.4 PROBLEMS
i. Demand analysis is challenging
ii. Response times are sluggish

An innovative smart grid concept has emerged to report the problems. SG is a vast
energy network that uses intelligent and real-time monitoring, communication, con­
trol, and self-healing technologies to ensure the reliability and security of clients’
electricity supply while offering a range of options. SGs are advanced cyber-physical
systems. The modern SG’s functionality can be divided into four categories.

1. Consumption: Energy is utilized by various industries and people for numerous

reasons.
2. Distribution: Power is disseminated to enable broader use.
3. Transmission: Electricity is transferred via a higher-voltage electronic
infrastructure.
4. Generation: Energy is generated using various techniques during this stage.

In the area of SG, ML, and DL functionalities have forecasting fraud detection,
optimal scheduling, stability of the SG, identification of security breaches, identifica­
tion of network anomalies, sizing, identification of faults, energy consumption, price,
and energy generation.

1.4.1 Computer Network

Important technical developments in networking, like network programmability
through Software-Defined Networking (SDN), have contributed to the utility of ML
in networking. Although ML has been applied widely to report problems, including
speech, and pattern recognition, its application in network processes and manage­
ment has been restricted. The main challenges are guessing what kind of data can
be assembled and what kind of controls can be executed on older network hardware.
These problems are mitigated by the SDN’s capacity to program the network. Routine
network administration and operation tasks are automated with the use of ML-based
cognition. It is, therefore, fascinating and problematic to apply ML methods to such
wide and complex networking problems. Because of this, ML in networking is an
intriguing field of study that calls for a deep comprehension of both networking and
ML methodologies.
Machine Learning: Advances and Applications 9

1.4.2 Energy System

A group of well-organized parts used for managing energy, manufacturing, and/or
transformation is called an energy system [22]. Mechanical, thermal, chemical, and
electro-magnetically elements might be united to form energy systems that cover a
broad range of energy-related categories, encompassing renewables and other types
of energy [22]. An enormous quantity of data grouped results in the increasing use
of data-gathering devices in power systems. Energy system development must make
difficult decisions to satisfy a range of exacting and competing objectives, including
budgetary constraints, ecological impact, effectiveness, and operational efficiency.
Smart sensors are progressively being used in the manufacture and utilization of
energy [23].
Making educated decisions is more problematic due to the abundance of oppor­
tunities presented by big data. The execution of big data technology in many uses has
been facilitated by using ML models. The energy industry has seen a rise in the usage
of prediction techniques based on ML models because they facilitate the inference
of functional correlations from observations [24]. ML models in energy systems are
essential for demand, making, and consumption predictive modeling because of their
speed, accuracy, and effectiveness. ML models provide an understanding of how the
energy system functions in the setting of intricate human relationships [25]. Both
conventional energy systems and substitute and renewable energy systems are created
using ML algorithms.

1.5 RECENT ADVANCES IN REAL-TIME USES

1.5.1 ML for ITS

Rapid car proliferation, inadequate transportation infrastructure, and the non­
existence of road security regulations are causing serious ecological and quality-
of-life issues in urban areas. For example, huge trucks frequently break the rules
of conventional highways in metropolitan areas, causing traffic jams and delays.
Furthermore, many bikers have frequented near-misses because of clothing, posture
changes, partial occlusions, and various observations that pointedly hinder the finding
rates of ML strategies. Enhancing the identification and classification of pedestrians,
cyclists, and special vehicles and license plate recognition (LPR) for a more sustain­
able and safer has garnered significant attention in the past ten years because of the
usage of machine learning (ML) and deep learning (DL) strategies for the evaluation
and presentation of a vast number of data collected from various sources. While a
great range of appearances can be captured by deep models, environment adapta­
tion is crucial. DL success is based on ANNs that mimic images by incorporating
interrelated node system groups that behave like the human brain. The nodes in a
nearby layer will be comprised of connections whose weights originate from nodes
in another layer. An instigation function in a node receives an input and a weight to
generate an output value.
10 IoT and Machine Learning for Smart Applications

1.5.2 ML for Healthcare

Activities and annotations are provided as input to policy functions throughout actions
carried out as response rewards. The technique acquired from this policy function is
called RL. RL is executed for various healthcare-based uses. Even RL is pragmatic
to the widespread symptom-checking method of disease detection. GOGAME is
another possible usage of RL [26].
When there are vast numbers of unlabeled and fewer labeled information access­
ible, SSL is appropriate. SSL uses both unlabeled and labeled data for training.
Applications for SSL in healthcare include medical picture segmentation [27],
activity recognition utilizing a variety of sensors suggested, and SSL for medical
care data clustering employed by the author [27]. Typical uses of SL in the medical
domain include nodule identification in pulmonary imaging and the recognition of
human parts utilizing several picture modalities.
USL involves training the model with unlabeled data to map input to
output: Clustering is based on (i) similarity, (ii) feature selection and dimensionality
reduction, and (iii) finding anomalies [28]. Various healthcare uses, including fea­
ture selection by Principal Component Analysis and heart disorder estimation using
clustering, can benefit from USL. Prognosis, finding, treatment, and medical workflow
are the four major fields of healthcare that can turnover from ML or DL techniques.

1.5.3 ML for Cybersecurity

In several industries like CS, design and manufacturing, medical [29], education, and
finance, AI and ML are extensively recognized and utilized. The subsequent domains of
CS - intrusion identification, dark or deeper websites, phishing, malware recognition,
fraud, and spam identification - all make extensive use of ML techniques. Vigorous and
innovative approaches are required to tackle the problems of CS as times change. Given
that ML is experience-based, it is amenable to evolutionary attacks. Classical CS tech­
nologies include firewalls, interruption preclusion systems, antivirus software, unified
threat management [30], and Security Information and Event Management solutions.
The author concluded that AI-based CS systems outperform classical CS systems
in terms of mistake rate, performance, and post-cyberattack response. According to
the study, cyberspace harm caused by an assault is only detected after it has occurred,
and this happens in about 60% of cases [31]. ML has a higher hold on both the
attacker and CS sides. Regarding CS, the goal is to defend everything from harm
created by attackers, identify attacks early on, and ultimately optimize performance.
On the invader’s side, ML is utilized to recognize holes and system vulnerabilities and
methods to become past firewalls. The study comes to the same conclusion regarding
how to improve classification performance by ML strategies.

1.5.4 ML for Computer Networks

1.5.4.1 Traffic Prediction
Traffic forecasting is crucial to the effective management and execution of network
operations since networks are growing more diverse and complicated every day,
Machine Learning: Advances and Applications 11

making it harder to administer and carry out network processes. Time series esti­
mating projects traffic patterns for the near future. Classifying network traffic is a
crucial step in managing and carrying out network operations. This process involves
resource provisioning, performance monitoring, and differentiation of services.

1.5.4.2 Congestion Control

Overflowing packets in a system will be regulated by a congestion controller. It
ensures that resource use is fair, network stability is maintained, and the packet loss
ratio stays within a satisfactory range.

1.6 EXISTING CHALLENGES TO ML TECHNOLOGY

ML is not without its tests, while technology shows potential and already serves
businesses worldwide. ML, for example, is good at identifying patterns but not so
good at generalizing information. Additionally, there is the problem of users becoming
weary of algorithms. In ML, a reasonable quantity of data and resources with good
performance are required for model training. Several Graphics Processing Units are
used to tackle this task. Real-time engineering uses require ML techniques that are
modeled to robustly solve a specific problem. It is essential to create a model specific­
ally for every task in real-time use since a single model created for one activity usage
cannot handle all the tasks across multiple domains.
Within the field of medicine, ML has the ability to anticipate illness incidence and
detect terrorist acts. As in Ribeiro et al. [32], relying solely on ML forecasts will not
prevent disastrous outcomes. While ML strategies are employed in many domains,
they demand extremely high degrees of correctness in certain domains as a substitute
for accuracy and speed. For a model to be considered trustworthy, it must be assured
that there is no shift in the dataset during training and testing. This was performed by
preventing data leaks [33].
The location of a moving object can be determined by permitting technologies,
including GPS and mobile phones; a crucial problem for ML is keeping this data safe
and unchangeable. The author claims that an object’s position data from numerous
sources are compared to identify similarities; the study corroborates the ambiguity in
the position data composed from numerous sources due to network delays and that
the veracity of such data desires to be addressed by ML strategies.
To facilitate trustworthy interactions among consumers and service providers in a
networked web system, an ontology of trust is put out [34]. Text classification also
makes use of trustworthiness. Trustworthiness combines both semantic and prac­
tical relations when interpreting the text’s meaning. It uses a metric model to verify
the software’s reliability. Like this study [35,36], the author claims that by using
ML strategies to create power-aware strategies, power usage in businesses and data
centers can be reduced. It’s advisable to try the machines off energetically to cut the
usage completely. The forecasting model will determine which machine should be
turned off; it is crucial to have faith in this model.
Certain problems associated with ML have important effects that are currently
being felt. The first is the “black box issue”, or the lack of interpretability and
explainability. Even the people who created it are baffled by how ML models come
12 IoT and Machine Learning for Smart Applications

to make their own decisions and actions. This makes it challenging to fix errors
and guarantee the impartiality and precision of a model’s output. For example,
when it was identified that women were given significantly smaller credit lines
by Apple’s credit card model than men, the company was not able to provide an
explanation or fix the issue. This relates to data and algorithmic bias, which are the
two biggest issues plaguing the area. Algorithms are regularly prejudiced against
women, Americans, and people of various civilizations. The topmost AI research
center in the world, Google DeepMind, has warned that LGBT people may be at
risk from the technology.
Even though this problem is widespread and well-known, there was oppos­
ition to the significant action that several industry professionals believe is required.
Researchers, legislators, and activists were surveyed, and the majority voiced concern
that as AI advances by 2030. The entire country is currently discussing and passing
laws about AI, especially regarding applications that are instantly and harmful recog­
nition for police enforcement.

1.7 UTILIZATIONS OF ML
In Industry 4.0, ML strategies are more common because of the capacity for intelli­
gent decision-making and for learning from the past. Here, we go over and condense
several application areas for ML algorithms.

1.7.1 Intelligent Decision-Making

ML strategies are applied to create intelligent decisions using data-driven [37]. The
foundation of predictive analytics is the capability to predict unknown results by
utilizing past events and identifying and catching the relationship between the fore­
cast and descriptive variables like credit card scams and criminal findings followed
by the crime. Prognostic analytics and intellectual decision-making are useful in
the retail sector for preventing out-of-stock situations, managing inventories, better
understanding customer behavior and preferences, and streamlining transportation
and warehousing. The most often utilized methods in the fields are ANN, decision
trees, and support vector machines [37]. Precisely forecasting the result can benefit
any firm, including those in social interaction, transport, sales and advertising,
healthcare, banking, financial facilities, telecommunication, E-commerce, and other
businesses.

1.7.2 Cyber-security and Hazard Intelligence

Cybersecurity (CS) is a key component of Industry 4.0 and is responsible for
safeguarding data, hardware, systems, and networks. One of the most significant
technologies in CS is ML, which protects users by encrypting cloud data, anticipating
malicious users online, identifying internal threats, and detecting malware in traffic.
CS and threat intelligence make use of DL-based security models, association rule
learning approaches, and ML categorization models [38].
Machine Learning: Advances and Applications 13

1.7.3 Smart Cities

By giving items the ability to send information and carry out tasks without the neces­
sity for human interaction, all objects in the Internet of Things (IoT) are transformed
into things. Business, healthcare, farming, retail, transportation, communique, edi­
fication, smart homes, smart governance, and smart cities [39] are a few industries
that use IoT. Because ML can evaluate data and forecast future events, it has emerged
as a key tool in the IoT. For example, in smart cities, congestion can be forecast,
judgments depend on the environment, energy can be estimated for a specific time,
and parking availability can be projected.

1.7.4 Sustainable Agriculture

All human endeavors depend on farming to thrive [40]. Sustainable farming practices
enhance crop yield while reducing adverse environmental effects. The authors [41,42]
clarify how novel technologies, mobile devices, and the IoT can be utilized to collect
vast amounts of information, which in turn can support the adoption of sustainable
agriculture techniques. Sustainable agriculture develops knowledge-intensive supply
chains using technology, skills, and information. The processing, production, and
preproduction phases of agriculture, as well as the distribution phases, can all benefit
from the application of various ML techniques. These include consumer study, inven­
tory control, production scheduling, estimating animal needs, managing soil nutrients,
identifying weeds and diseases, forecasting the weather, determining irrigation needs,
analyzing soil characteristics, and projecting crop yields.

1.7.5 COVID-19 Pandemic and Healthcare

ML may help address diagnostic and prognostic challenges in a range of medically
relevant application fields, such as sickness prediction, medical information extrac­
tion, data regularity detection, patient data management, etc. [43]. The WHO classifies
coronavirus as an infectious disease. In the fight against COVID-19, learning strat­
egies have recently gained prominence. The COVID-19 pandemic’s death rate, high-
risk patients, and other anomalies are being categorized using learning techniques. It
is utilized to its fullest to recognize the source of the virus, forecast the COVID-19
outbreak, and identify and manage the illness [44]. ML may be used by researchers
to forecast the locations and times at which COVID- 19 would spread, allowing those
areas to be prepared. DL can offer more effective solutions to the therapeutic image
processing issues related to the COVID-19 epidemic. All things considered, DL and
ML processes can aid in fighting the coronavirus and epidemic, possibly even helping
create smart clinical decisions in the medical field.

1.7.6 Pattern Recognition

The detection of objects in images is the goal of image recognition, which makes exten­
sive use of ML [45]. A few examples of image recognition are the cancer label on
an X-ray image, face detection, character recognition, and social media suggestion
14 IoT and Machine Learning for Smart Applications

tagging. The well-known language and sound models in speech recognition are Alexa,
Siri, Cortana, Google Assistant, etc. Pattern recognition is the automatic identification
of regularities and patterns in data, like image study [45]. Various ML strategies are
used in this field, like feature selection, classification, clustering, and sequence labeling.

1.8 TRIALS AND FUTURE GUIDELINES IN ML

This chapter examines the pertinence of various ML algorithms in the study of
intelligent data and applications, which raises several research concerns. Research
prospects and possible future initiatives are outlined and explored here.
The overview of research directions:

i. It’s crucial to focus on the thorough examination of data-collecting methods

while working with real-world data. It is important to enhance existing
preprocessing techniques or create new ones to manage real-world data
associated with usage domains.
ii. One of the research interests is determining which ML technique is best for
the intended application.
iii. There is a great deal of interest in academia in proposing new hybrid
algorithms and enhancing or modifying existing ML algorithms to make them
more applicable to the domain of target applications.

The efficacy and efficiency of ML solutions are determined by the nature as well as
the qualities of the information, along with the performance of ML processes over it.
The development of vast volumes of data in a short period by numerous application
domains, like cyber-security, healthcare, as well as agriculture, makes data collecting
in these areas challenging. Relevant data collecting is essential to moving on with the
investigation of the data in ML-based uses. Consequently, while working with real-
world information, attention must be paid to a deeper analysis of the data collection
techniques.
We must do the challenging process of purifying the acquired data from numerous
sources because there might be numerous outliers, missing and confusing values in
the prevailing information, which will impair the training of ML strategies. Therefore,
to maximize the utilization of ML strategies, pretreatment should be enhanced, and a
novel preprocessing procedure should be used.
The features and nature of the information might affect the results of the various
ML strategies, making it difficult to choose the ML procedure that is better opted for
the target use, for removing understandings, and for data analysis. An improper ML
process will provide unforeseen results, which could lessen the model’s efficacy and
accuracy. These algorithms are adjusted for the intended usage domains, or novel
techniques must be suggested.
The uses and their associated ML-based solutions’ final success will be based on
ML algorithms and the characteristics of the data. When the data is not representa­
tive, contains irrelevant characteristics, is of low quality, or is not enough for training,
ML models will produce fewer accurate results and become unusable. Two important
Machine Learning: Advances and Applications 15

things are needed to create an intelligent application: Managing different learning

strategies and efficiently processing data.
Many new questions in the field of ML methods for intelligent data and applications
are brought up by our study. Consequently, we highlight the challenges discussed in
this section together with potent avenues for upcoming study and activities. The effi­
cacy and efficiency of an ML-based solution are based on the type and quality of the
information as well as the functioning of the learning procedures. To compile data in
a particular field, like IoT, CS, healthcare, farming, etc. Data is thus gathered for the
intended ML-based applications. It is necessary to thoroughly analyze data-gathering
procedures while working with real-world information. Moreover, there may be a
significant amount of missing data, outliers, unclear values, and meaningless data in
historical data.
Numerous ML algorithms are available to evaluate data and derive perceptions,
but both the data and the learning procedures play a major part in the overall perform­
ance of an ML-based solution and its supporting applications. Reduced accuracy will
result from data that was critical to learn from, like non-representative, low-quality,
inappropriate features, or not enough of it for training. Because of this, it’s essential
to develop an ML-based solution, design intelligent apps, process data accurately, and
manage different learning algorithms.

1.9 CONCLUSION
This study on ML techniques includes an extensive examination of data analysis.
Here, a summary of real-world problems and how various learning processes are
employed to produce answers is provided. The performance of ML procedures and
the properties of data will determine the accomplishment of the ML model. ML
algorithms must be educated by data collected from diverse real-world scenarios and
familiar with target application knowledge to produce intelligent decision-making.
In this overview, a range of application areas and challenges in the real world are
highlighted about the applicability of ML techniques. Finally, a summary and discus­
sion of future research prospects and other problems are provided. Every problem in
the target uses field needs to be solved with the help of efficient solutions. This chapter
will function as a sample point for experts in the industry as well as academia. From
the methodological viewpoint, it also serves as a yardstick for decision-makers across
various usage realms followed by real-world situations. The applications of ML are
not limited to a single industry. Instead, it is affecting various areas, like gaming,
broadcasting and entertainment, information technology, lending and finance, and the
car industry. Because ML has so many applications, researchers are working in sev­
eral fields to attempt and change the nation.

REFERENCES
1. Sharma, D. and Kumar, N., 2017. A review on machine learning algorithms, tasks and
applications. International Journal ofAdvanced Research in Computer Engineering &
Technology (IJARCET), 6(10), pp. 215-223.
16 IoT and Machine Learning for Smart Applications

2. Sandhu, T.H., 2018. Machine learning and natural language processing - a

review. International Journal of Advanced Research in Computer Science, 9(2),
pp. 582-584.
3. Sarker, I.H., Furhad, M.H. and Nowrozy, R., 2021. AI-driven cybersecurity: an over­
view, security intelligence modeling and research directions. SN Computer Science, 2,
pp. 1-18.
4. Sarker, I.H., Hoque, M.M., Uddin, M.K. and Alsanoosy, T., 2021. Mobile data science
and intelligent apps: concepts, AI-based modeling and research directions. Mobile
Networks and Applications, 26, pp. 285-303.
5. Sarker, I.H., Kayes, A.S.M., Badsha, S., Alqahtani, H., Watters, P. and Ng, A., 2020.
Cybersecurity data science: an overview from machine learning perspective. Journal
of Big Data, 7, pp. 1-29.
6. Slusarczyk, B., 2018. Industry 4.0 - are we ready? Polish Journal of Management
Studies, 17(1), pp. 232-248.
7. Zhu, X. and Goldberg, A.B., 2020. Introduction to semi- supervised learning
(Vol. 109, pp. 373-440). Cham, Switzerland: Springer Nature.
8. Sutton, R.S., 1992. Introduction: the challenge of reinforcement learning. In Richard
S. Sutton (ed.), Reinforcement learning (pp. 1-3). Boston, MA: Springer US.
9. Caruana, R., 1997. Multitask learning. Machine Learning, 28, pp. 41-75.
10. Opitz, D. and Maclin, R., 1999. Popular ensemble methods: an empirical study.
Journal ofArtificial Intelligence Research, 11, pp. 169-198.
11. Sharma, V., Rai, S. and Dev, A., 2012. A comprehensive study of artificial neural
networks. International Journal of Advanced Research in Computer Science and
Software Engineering, 2(10), pp. 278-284.
12. Hiregoudar, S.B., Manjunath, K. and Patil, K.S., 2014. A survey: research summary
on neural networks. International Journal ofResearch in Engineering and Technology,
3(15), pp. 385-389.
13. Bhattacharyya, D.K. and Kalita, J.K., 2013. Network anomaly detection: A machine
learning perspective. Boca Raton, FL: CRC Press.
14. Ambalavanan, V., 2020. Cyber threats detection and mitigation using machine learning.
In T. Thomas, A. P. Vijayaraghavan, and S. Emmanuel (eds.), Handbook of research
on machine and deep learning applications for cyber security (pp. 132-149). Hershey,
PA: IGI Global.
15. Thomas, T., Vijayaraghavan, A.P. and Emmanuel, S., 2020. Machine learning
approaches in cyber security analytics (pp. 37-200). Singapore: Springer.
16. Latif, S., Qadir, J., Farooq, S. and Imran, M.A., 2017. How 5G wireless (and concomi­
tant technologies) will revolutionize healthcare?. Future Internet, 9(4), p. 93.
17. Havaei, M., Davy, A., Warde-Farley, D., Biard, A., Courville, A., Bengio, Y., Pal, C.,
Jodoin, P.M. and Larochelle, H., 2017. Brain tumor segmentation with deep neural
networks. Medical Image Analysis, 35, pp. 18-31.
18. Latif, S., Asim, M., Usman, M., Qadir, J. and Rana, R., 2018. Automating motion
correction in multishot MRI using generative adversarial networks. 32nd Conference
on Neural Information Processing Systems (NIPS 2018), Montreal, Canada. arXiv pre­
print arXiv:1811.09750.
19. Mao, Q., Hu, F. and Hao, Q., 2018. Deep learning for intelligent wireless
networks: a comprehensive survey. IEEE Communications Surveys & Tutorials, 20(4),
pp. 2595-2621.
20. Deshmukh, M.K. and Deshmukh, S.S., 2008. Modeling of hybrid renewable energy
systems. Renewable and Sustainable Energy Reviews, 12(1), pp. 235-249.
Machine Learning: Advances and Applications 17

21. Mosavi, A., Salimi, M., Faizollahzadeh Ardabili, S., Rabczuk, T., Shamshirband, S.
and Varkonyi-Koczy, A.R., 2019. State of the art of machine learning models in energy
systems, a systematic review. Energies, 12(7), p. 1301.
22. Smets, A., Jager, K., Isabella, O., Van Swaaij, R. and Zeman, M., 2016. Solar
energy: the physics and engineering of photovoltaic conversion, technologies and
systems. London: Bloomsbury Publishing.
23. Torabi, M., Hashemi, S., Saybani, M.R., Shamshirband, S. and Mosavi, A.,
2019. A Hybrid clustering and classification technique for forecasting short­
term energy consumption. Environmental Progress & Sustainable Energy, 38(1),
pp. 66-76.
24. Yildiz, B., Bilbao, J.I. and Sproul, A.B., 2017. A review and analysis of regression
and machine learning models on commercial building electricity load forecasting.
Renewable and Sustainable Energy Reviews, 73, pp. 1104-1122.
25. Abidi, M.H., Alkhalefah, H., Moiduddin, K., Alazab, M., Mohammed, M.K., Ameen,
W. and Gadekallu, T.R., 2021. Optimal 5G network slicing using machine learning and
deep learning concepts. Computer Standards & Interfaces, 76, p. 103518.
26. Kao, H.C., Tang, K.F. and Chang, E., 2018, April. Context-aware symptom checking
for disease diagnosis using hierarchical reinforcement learning. In Proceedings of the
AAAI conference on artificial intelligence (vol. 32, no. 1), pp. 1-7.
27. Sohail, M. N., Ren, J. and Uba Muhammad, M., 2019. A Euclidean group assessment
on semi-supervised clustering for healthcare clinical implications based on real-life
data. International Journal of Environmental Research and Public Health, 16(9),
p. 1581.
28. Ramana, K., Ponnavaikko, M. and Subramanyam, A., 2019. A global dispatcher load
balancing (GLDB) approach for a web server cluster. In ICCCE 2018: proceedings of
the international conference on communications and cyber physical engineering 2018
(pp. 341-357). Springer Singapore.
29. Jha, S. and Topol, E.J., 2016. Adapting to artificial intelligence: Radiologists and
pathologists as information specialists. JAMA, 316(22), pp. 2353-2354.
30. Zhang, X. and Chow, K.P., 2020. A framework for dark web threat intelligence analysis.
In Cyber warfare and terrorism: concepts, methodologies, tools, and applications (pp.
266-276). Hershey, PA: IGI Global.
31. Sharma, A., Kalbarczyk, Z., Barlow, J. and Iyer, R., 2011, June. Analysis of security
data from a large computing organization. In 2011IEEE/IFIP 41st international con­
ference on dependable systems & networks (DSN) (pp. 506-517). IEEE.
32. Ribeiro, M.T., Singh, S. and Guestrin, C., 2016, August. “Why should I trust you?”
Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD
international conference on knowledge discovery and data mining, San Francisco,
California, pp. 1135-1144.
33. Ghosh, S., Lincoln, P., Tiwari, A. and Zhu, X., 2017, March. Trusted machine
learning: model repair and data repair for probabilistic models. In Workshops at
the thirty-first AAAI conference on artificial intelligence, San Francisco, California,
pp. 143-152.
34. Zhu, M. and Jin, Z., 2009, September. A trust measurement mechanism for service
agents. In 2009 IEEE/WIC/ACM international joint conference on web intelligence
and intelligent agent technology (vol. 1, pp. 375-382). New York, NY: IEEE.
35. Tao, H. and Chen, Y., 2009, September. A metric model for trustworthiness of
softwares. In 2009 IEEE/WIC/ACM international joint conference on web intelligence
and intelligent agent technology (vol. 3, pp. 69-72). New York, NY: IEEE.
18 IoT and Machine Learning for Smart Applications

36. Berral, J.L., Goiri, I., Nou, R., Julia, F., Guitart, J., Gavalda, R. and Torres, J., 2010,
April. Towards energy-aware scheduling in data centers using machine learning.
In Proceedings of the 1st international conference on energy- efficient computing
and networking (pp. 215-224). New York, NY: ACM (Association for Computing
Machinery).
37. Li, X.H., Cao, C.C., Shi, Y., Bai, W., Gao, H., Qiu, L., Wang, C., Gao, Y., Zhang, S.,
Xue, X. and Chen, L., 2020. A survey of data-driven and knowledge-aware explainable
AI. IEEE Transactions on Knowledge and Data Engineering, 34(1), pp. 29-49.
38. Sarker, I.H., Abushark, Y.B., Alsolami, F. and Khan, A.I., 2020. IntruDTree: a machine
learning based cyber security intrusion detection model. Symmetry, 12(5), p. 754.
39. Zanella, A., Bui, N., Castellani, A., Vangelista, L. and Zorzi, M., 2014. Internet of
things for smart cities. IEEE Internet of Things Journal, 1(1), pp. 22-32.
40. Adnan, N., Nordin, S.M., Rahman, I. and Noor, A., 2018. The effects of knowledge
transfer on farmers decision making toward sustainable agriculture practices: In view
of green fertilizer technology. World Journal of Science, Technology and Sustainable
Development, 15(1), pp. 98-115.
41. Kamble, S.S., Gunasekaran, A. and Gawankar, S.A., 2018. Sustainable Industry 4.0
framework: A systematic literature review identifying the current trends and future
perspectives. Process Safety and Environmental Protection, 117, pp. 408-425.
42. Kamble, S.S., Gunasekaran, A. and Gawankar, S.A., 2020. Achieving sustainable
performance in a data-driven agriculture supply chain: a review for research and
applications. International Journal of Production Economics, 219, pp. 179-194.
43. Ardabili, S.F., Mosavi, A., Ghamisi, P., Ferdinand, F., Varkonyi-Koczy, A.R., Reuter,
U., Rabczuk, T. and Atkinson, P.M., 2020. COVID-19 outbreak prediction with
machine learning. Algorithms, 13(10), p. 249.
44. Alakus, T.B. and Turkoglu, I., 2020. Comparison of deep learning approaches to pre­
dict COVID-19 infection. Chaos, Solitons & Fractals, 140, p. 110120.
45. Chiu, C.C., Sainath, T.N., Wu, Y., Prabhavalkar, R., Nguyen, P., Chen, Z., Kannan, A.,
Weiss, R.J., Rao, K., Gonina, E. and Jaitly, N., 2018, April. State-of-the-art speech rec­
ognition with sequence-to-sequence models. In 2018 IEEE international conference
on acoustics, speech and signal processing (ICASSP) (pp. 4774-4778). New York,
NY: IEEE.
 Understanding the
2 Concept of IoT
C. Emilin Shyni, Megha Menon K.,
and Anesh D. Sundar ArchVictor

2.1 INTRODUCTION
Imagine waking up to the sound of your favorite melody, with the curtains in your
bedroom gently pulling back to reveal the first rays of sunshine, your coffee machine
already brewing the perfect cup to jumpstart your day. This isn’t a scene from a futur­
istic movie; it’s a slice of the present, brought to life by the Internet of Things (IoT).
A silent revolution has unfurled throughout homes and cities around us, connecting
devices and everyday objects to work seamlessly, making our lives more comfortable
and efficient. But what exactly lies behind this technological evolution that’s subtly
yet persistently transforming our daily routines?
IoT is no longer a buzzword confined to tech enthusiast circles; it has become an
integral part of our lives, and its relevance is proliferating with each passing day. This
intricate web of interconnected gadgets and sensors offers a symphony of conveni­
ence that orchestrates our thermostats, vehicles, and even healthcare devices. It’s a
concept that weaves itself into the very fabric of our existence, aiming to optimize
our world by enhancing communication between devices - as much as we humans
communicate with each other.
Yet, despite its ubiquitous nature, the essence of IoT is shrouded with a sense of
mystique for many.
For the uninitiated, the IoT can seem overwhelming - a complex network of intel­
ligent objects turning the mundane into the magical. But fear not, for understanding
IoT doesn’t require a degree in rocket science; it simply demands curiosity and a
willingness to embrace the future. As we usher in an era where your car notifies your
heater to warm up the house in anticipation of your arrival, isn’t it time to demystify
the magic and grasp the strings that control this technological puppetry?
This blog post aims to pull back the curtain on IoT, revealing the wizardry of
interconnected tech in a language you’ll not only comprehend but relate to. We’re
about to dive into the sprawling ecosystem of the IoT to unveil how it shapes our
modern lives and, more importantly, how it stands to revolutionize the future. Ready
to embark on a journey of discovery that will change the way you interact with the
world around you? Then let’s get connected, for this is a narrative that stretches from
the smallest microchip to the largest cities, narrating a story where you are an insep­
arable protagonist.

DOI: 10.1201/9781032623276-2 19
20 IoT and Machine Learning for Smart Applications

2.2 DEFINITION OF IOT

To put it simply, the IoT is a network of linked objects that are capable of exchan­
ging data and interacting with one another. These gadgets range from medical
equipment and animals to industrial gear, transit systems, cell phones, and home
appliances.
Increasing the intelligence and connectivity of common things to enable data
transmission and reception, remote control, and interaction with other systems
and devices is the aim of the IoT. This has the potential to completely trans­
form the way we work and live by enabling innovative and effective methods
of operation and by giving governments and corporations access to insightful
information.

2.3 HISTORICAL CONTEXT AND EVOLUTION OF IOT

Are you aware that the idea of the IoT has existed for some time? You might be
shocked to hear that Kevin Ashton, the “father of IoT”, originally used the word in
1998 while he was employed by Procter & Gamble. When smartphones and tablets
became extensively used in 2011 and 2012, the IoT began to receive attention in the
media. These days, Industry 4.0, smart cities, and smart homes are just a few areas in
which the IoT is included. Part 1 of this two-part chapter will examine the evolution
of IoT and its historical context.

Historical Context:
The concept of IoT was first coined in the early 1990s when researchers started
exploring the possibility of connecting everyday objects to networks. However, the
term “Internet of Things” was coined in 1998 by Kevin Ashton, a British entrepreneur
who worked at Procter & Gamble.

Evolution of IoT:

1. Early Stage (1990s-2000s):

In the early stages, IoT research focused on connecting everyday objects to
networks. However, it wasn’t until the early 2000s that the idea of IoT gained
mainstream attention.
2. Growth Stage (2010-2012):
The widespread adoption of smartphones and tablets led to the creation of IoT
applications. Companies started developing IoT-enabled devices that could be
controlled and monitored through mobile devices.
3. Integration Stage (2013-2018):
With the increase in IoT devices, companies began integrating IoT into various
industries, including smart homes, healthcare, and agriculture.
4. Maturity Stage (2019-Today):
The maturity stage saw the integration of IoT into various aspects of our lives,
from smart homes to smart cities and Industry 4.0.
Understanding the Concept of IoT 21

Future of IoT:
From 2020 to 2027, the IoT market is anticipated to expand at a compound annual
growth (CAGR) of 25.2%. The rising need for new IoT applications, such as linked
automobiles and smart factories, is the reason for this rise.
The concept of IoT has been around for decades, and over time, it has evolved to
become an integral part of our lives. From smart homes to smart cities and Industry 4.

2.4 IMPORTANCE AND CURRENT RELEVANCE

The IoT is a very hot topic right now. If you’re not familiar with it, the concept refers
to everyday devices being connected to the internet and controlled remotely.
IoT products are used in homes (such as smart thermostats, security cameras, and
baby monitors), businesses, and factories. The IoT trend will grow as technology
improves and more everyday objects become connected.
Some major players in IoT are Amazon, Google, and Apple. These companies are con­
stantly developing new ways to connect everyday items over the Internet. For example:
Siri voice commands are used in Apple’s HomeKit to control their smart home
devices.
Amazon’s Echo connects to various smart home devices, allowing users to control
them using voice commands.
Google is currently developing its own smart home platform called Google Home.

2.5 IOT TECHNOLOGIES AND ARCHITECTURE

An ecosystem of equipment and gadgets that are linked to the Internet and can com­
municate with one another is referred to as the IoT.
Data from IoT devices is communicated, examined, and used to make decisions.
Sensors, data storage, communication protocols, and applications make up the layers
of an IoT architecture.
IoT architecture includes three layers:

Sensors and devices: These gather information from the real world and transmit it
to the IoT platform.
Protocols for communication: Data is sent between devices and the cloud via these
protocols. The most widely used protocols are Bluetooth Low Energy (BLE),
Wi-Fi, and Zigbee.
Data processing and storage: The databases and analytical tools that make up
this layer are used to store and handle the data that IoT devices have acquired.
Additionally, it offers machine-generated insights and user interfaces for
decision-making.

2.5.1 Sensors and Actuators

Sensors and actuators are key components in various systems, particularly in the fields
of engineering, automation, and control. They play an important role in collecting and
processing data, as well as driving physical actions in response to that data.
22 IoT and Machine Learning for Smart Applications

Sensors are instruments that are capable of detecting and quantifying various phys­
ical events or attributes, including motion, light, humidity, pressure, and temperature.
They transform the signals they detect into electrical signals or digital data that other
system components can process with ease. It is usual to find temperature, pressure,
motion, proximity, and ultrasonic sensors, among other sorts of sensors.
Actuators, on the other hand, are devices that convert electrical signals or digital
commands into physical actions. They are responsible for controlling and moving
physical systems, such as opening and closing valves, adjusting motor speeds, or
operating robotic arms. Examples of actuators include motors, solenoids, relays,
pumps, and pneumatic or hydraulic cylinders.
In many applications, sensors and actuators work together in a closed-loop
system. Sensors detect and measure certain physical parameters, and the obtained
data is processed by a controller or a computer. The controller then sends appropriate
commands to actuators to perform specific actions based on the collected data. This
feedback loop allows systems to continuously monitor and adapt to changes in their
environment, ensuring efficient and accurate operation.
Numerous industries and applications, including robotics, aircraft, automo­
tive systems, industrial automation, healthcare, and smart homes, use sensors and
actuators. They make it possible to create intelligent, self-governing systems that
can communicate with the outside world and react instantly to changes in their
surroundings.

2.5.2 Connectivity and Network Protocols

The capacity of various networks and devices to connect and communicate with
one another is referred to as connectivity. In order to facilitate data interchange and
transmission, it entails establishing a steady and dependable link between devices or
networks. Figure 2.1 shows the IoT protocols.
The sort of data being transmitted and the particular network requirements deter­
mine which protocols are used.

2.5.3 Edge and Cloud Computing

While cloud computing and edge computing are closely related, they have various
uses in the computing industry.
The process of processing and evaluating data at or close to the location where it is
generated, as opposed to sending it to a central data center or the cloud, is known as
edge computing. Real-time data processing is made possible by this method, which
lowers latency and speeds up reaction times. Edge computing is especially helpful
in situations when network bandwidth is limited, including in remote or mobile
contexts, or when low latency is essential, like in industrial IoT (IIoT) applications or
autonomous vehicles.
On the other hand, cloud computing involves using a network of remote servers
to store, manage, and process data rather than utilizing local servers or personal
computers. This allows for scalable and on-demand access to computing resources,
enabling organizations to easily provision and deploy applications and services
Understanding the Concept of IoT 23

FIGURE 2.1 IOT protocols.

without having to invest heavily in infrastructure. Cloud computing provides high

availability, resilience, and flexibility for various workloads and allows data to be
accessed from anywhere with an Internet connection.
Edge computing and cloud computing are complementary to each other and can
work in tandem to provide a comprehensive computing solution. Edge devices can
collect and process data at the edge, while the cloud can provide storage, data ana­
lysis, and management capabilities. This combined approach is known as fog com­
puting, where a hybrid architecture of edge and cloud resources is used to efficiently
process and manage data.
Overall, edge computing focuses on decentralized data processing and real-time
decision-making, while cloud computing provides scalable and centralized com­
puting resources. Each strategy has benefits of its own and can be applied depending
on the particular needs of a use case or application.

2.6 IOT PLATFORM AND MIDDLEWARE

The creation and implementation of IoT solutions require IoT platforms and middle­
ware. They offer the infrastructure, services, and tools required to link and control the
different systems and gadgets in an IoT network.
24 IoT and Machine Learning for Smart Applications

IoT platforms serve as a central hub for collecting, processing, and analyzing the
data generated by IoT devices. They typically offer functionalities such as device
management, data management, real-time analytics, and integration with other
systems or applications. These platforms enable organizations to monitor and control
their IoT devices, as well as derive meaningful insights from the collected data.
Middleware, on the other hand, acts as a bridge between the IoT devices and the
applications or services that utilize the data. It provides services such as data trans­
formation, protocol conversion, message routing, and security. Middleware effect­
ively manages the communication and interaction between IoT devices and backend
systems, allowing seamless integration and interoperability.

2.7 IOT COMPONENTS AND ECOSYSTEM

The IoT ecosystem consists of various interconnected components, including hard­
ware devices, communication protocols, network protocols, APIs, and software
frameworks. These devices and protocols enable IoT devices to communicate with
one another and exchange data over the Internet.
Message Queuing Telemetry Transport (MQTT), Constrained Application Protocol
(CoAP), and Extensible Messaging and Presence Protocol (XMPP) are a few of the
widely used communication protocols in the IoT ecosystem. These protocols allow
IoT devices to communicate with servers, gateways, and other IoT devices over
various types of networks, including Wi-Fi, cellular, and satellite networks.

2.8 DEVICES AND HARDWARE

Some common devices and hardware in IoT include:

1. Sensors: Various factors, including temperature, humidity, light, pressure,

motion, and more, can be measured and collected using these devices. Data
from the physical world is captured in real-time with the use of sensors.
2. Actuators: Actuators are devices that enable physical actions or changes based
on the data received from sensors. They convert electrical signals into mech­
anical movements, such as opening/closing valves, turning switches on/off,
adjusting positions, etc.
3. Microcontrollers: These are small computer chips that integrate micropro­
cessor, memory, and programmable input/output peripherals. Microcontrollers
are often used to control and manage IoT devices, enabling them to process
data and execute commands.
4. Communication modules: These modules facilitate wireless connectivity,
allowing IoT devices to communicate with each other and with the central
server or cloud-based platforms. Common communication protocols used in
IoT include Wi-Fi, Bluetooth, Zigbee, cellular networks (2G, 3G, 4G, and
5G), and LoRaWAN.
5. Embedded systems: Embedded systems are hardware and software components
integrated into a larger IoT device or system. They are designed for specific
Understanding the Concept of IoT 25

functions and are often low-power, resource-constrained, and specialized for

a particular application.
6. Cloud-based platforms: While not physical devices, cloud-based platforms
are an essential component of IoT infrastructure. They provide storage, com­
puting power, and analytics capabilities, enabling the processing and ana­
lyzing vast amounts of IoT data.
7. Robotics and automation systems: IoT-enabled robots and automated systems
use sensors, actuators, and connectivity to perform tasks autonomously or
with minimal human intervention. These devices find applications in indus­
tries like manufacturing, healthcare, agriculture, logistics, etc.
8. Smart home devices: Smart appliances, door locks, security cameras, lighting
controls, and thermostats are examples of IoT devices that are part of the smart
home. They may be operated remotely with voice assistants or smartphones
by connecting them to a home network.

Overall, the devices and hardware used in IoT systems enable the collection, trans­
mission, processing, and control of data, making IoT applications possible across
various domains.

2.9 DATA ANALYTICS AND AI IN IOT

1. Anomaly Detection: Data analytics tools and AI algorithms help in detecting
anomalies in the IoT data. Unusual patterns or deviations from normal
behavior can indicate potential issues, security breaches, or anomalies in the
system. Early detection of anomalies can prevent potential failures or threats
to the IoT infrastructure.
2. Energy Optimization: With the help of data analytics and AI, IoT devices can
optimize energy consumption. By analyzing data on energy usage patterns,
organizations can identify inefficiencies and implement measures to optimize
energy consumption, reduce costs, and improve sustainability.
3. Personalization and User Experience: By analyzing user behavior and
preferences, IoT devices combined with AI offer personalized experiences.
With data analytics, organizations can derive insights about user preferences,
habits, and choices, enabling them to deliver personalized services and
enhance the overall user experience.
4. Cybersecurity: IoT devices are prone to cybersecurity threats due to their
interconnected nature. AI and data analytics can help identify and mitigate
potential security breaches or anomalies. Machine learning algorithms can
detect abnormal patterns in data traffic and alert system administrators.

Data analytics and AI are critical components in leveraging the full potential of IoT.
By collecting, analyzing, and interpreting the enormous amount of data generated
by IoT devices, organizations can optimize operations, improve decision-making,
enhance user experience, and ensure the security of their IoT ecosystem.
26 IoT and Machine Learning for Smart Applications

2.10 SECURITY AND PRIVACY CONSIDERATIONS

You should also consider buying a security camera that encrypts your video feed
while it’s being transmitted. That’s especially important if you have neighbors or
others who can intercept your Wi-Fi signal.
Additionally, you should only share your camera’s video feed with people you
trust. If you have an outdoor camera pointed at your front door, don’t share that link
with strangers via social networks or messaging apps.
If you want more security and privacy, consider buying a security camera that
stores its footage locally on an SD card or a hard drive rather than in the cloud. That’s
the only way to guarantee that your footage won’t be accessed by hackers or other
people without your permission.

2.11 IOT IN DAILY LIFE

With its incredible characteristics, the IoT is changing the world. It is a network of
actual physical objects, including buildings, cars, and other things, that are linked
to the Internet and can gather and exchange data. IoT devices are capable of com­
municating, sharing data, and carrying out activities without the need for human
assistance.
IoT is applied in several industries, including home automation, transportation,
healthcare, and agriculture. Here are a few instances of IoT applications in daily life:

1. Smart houses: To improve comfort, efficiency, and energy conservation, smart

houses employ IoT devices. For instance, smart lighting may automatically
adapt based on occupancy and preferences, and smart thermostats can be con­
trolled remotely.
2. Health Monitoring: Biometric data, such as heart rate, blood pressure, and
sleep habits, are tracked by IoT-enabled health monitoring devices.
3. Vehicle Telematics: IoT devices are utilized in automobiles to offer functions, such
as emergency assistance, vehicle tracking, and remote locking and unlocking.
4. Smart Agriculture: IoT devices are used to monitor cattle, crops, and soil
conditions.

2.11.1 Smart Homes and Consumer Applications

A smart home is an automated system that uses Internet-based technology to manage
the lights, appliances, temperature, and other elements of a house. The idea behind a
smart home is to combine several gadgets into a single, controllable system that can
be accessed through a mobile app or central hub.
The concept of a smart home is relatively new, and many homeowners are excited
about its potential to enhance their lifestyle. However, the concept of a smart home is
not limited to just residential use.
A smart home can also be applied to the commercial sector. For example,
businesses can use smart technology to control various aspects of their operations,
such as energy usage, security, and employee productivity.
Understanding the Concept of IoT 27

2.11.2 Wearables and Personal IOT Devices

Wearables and personal IoT devices have been making headlines in recent years
and are becoming increasingly popular. These devices are Internet-enabled and can
connect to the Internet using WiFi or cellular networks.
Smartwatches, fitness trackers, and augmented reality glasses are examples of
wearable technology that is intended to be worn on the body to gather information
about an individual’s activities and bodily processes.
Devices classified as personal IoT are those that are usually not wearable and
intended for usage in certain settings or contexts.
These gadgets can be personal medical equipment like blood pressure cuffs, glu­
cose monitors, and heart rate monitors, as well as smart home gadgets like security
cameras, smart lighting, and smart thermostats.
Wearables and personal IoT devices have a lot of advantages. Users who use these
gadgets can increase productivity, maintain relationships with friends and family, and
lead healthier lives.

2.11.3 Smart Cities and Urban Development

There are several key areas in which smart cities focus on urban development:

1. Infrastructure: Smart cities invest in advanced infrastructure to support sus­

tainable and efficient operations. This includes using smart grids for elec­
tricity distribution, intelligent transportation systems for traffic management,
and integrated sensor networks for monitoring and managing utilities such as
water supply and waste management.
2. Mobility: Smart cities aim to offer sustainable and effective transit
choices. To lessen traffic, improve air quality, and increase accessibility
for all citizens, they use technology, including shared mobility services,
intelligent traffic management systems, and real-time public transporta­
tion tracking.
3. Public safety: Modern technologies are used in smart cities to improve
security and safety for the general public. This includes employing real-time
emergency response systems for quicker assistance during emergencies, pre­
dictive analytics to pinpoint crime hotspots, and facial recognition technology
in video surveillance systems.
4. Environment: Smart cities prioritize sustainability and environmental con­
servation. They deploy smart energy grids, implement renewable energy
sources, and promote energy-efficient buildings. They also use data
analytics to monitor air quality, reduce waste generation, and improve
water management.
5. E-Governance: Smart cities leverage technology to enhance citizen engage­
ment and improve the delivery of government services. They develop online
platforms and mobile apps for citizens to access information, make payments,
and provide feedback. These cities also use data analytics to identify patterns
and make data-driven policy decisions.
28 IoT and Machine Learning for Smart Applications

Urban development and smart cities work hand in hand to build livable, sustain­
able, and resilient urban environments through efficient use of data and tech­
nology. Smart cities improve the quality of life for their citizens by integrating
e-governance techniques, public safety measures, mobility solutions, and modern
infrastructure.

2.11.4 Internet of Things in Industry

The sector could undergo a change thanks to the IoT, which could make asset
tracking, process monitoring, and operational optimization easier and more afford­
able for businesses.

2.11.5 Industrial IOT (IIOT) and Manufacturing

Businesses are using IoT technologies to increase productivity and efficiency as they
continue to automate and modernize their processes.
IIoT and manufacturing are two areas where IoT is having a big influence on
industry.
IIoT refers to the process of improving operations and boosting production by
connecting industrial equipment, such as machines, sensors, and control systems,
to the Internet. Businesses may have real-time visibility into their operations by
gathering and evaluating data from these devices. This allows them to optimize their
procedures and make data-driven decisions.
IoT technologies are especially well-suited to the manufacturing sector, which
uses a variety of sophisticated machinery and processes that may be made more effi­
cient by gathering and analyzing data in real time. Manufacturers, for instance, can
employ IoT-enabled devices.

2.11.6 Agriculture and Environmental Monitoring

IoT in agriculture and environmental monitoring allows for the remote monitoring
and analysis of crops, soil, and environmental conditions. IoT devices such as sensors
placed on agricultural equipment, IoT devices in soil, and IoT devices in the atmos­
phere collect real-time data that farmers, researchers, and policymakers can use to
make informed decisions.
IoT devices in agriculture and environmental monitoring can also be used to
monitor air and water quality, which is essential for maintaining soil and water quality
and protecting the environment. These devices can be used to identify and measure
levels of pollutants, such as nitrogen and phosphorus, in water and soil, which can
contribute to algal blooms and eutrophication of water bodies.
Overall, IoT devices in agriculture and environmental monitoring can help
farmers, researchers, and policymakers make informed decisions that improve crop
yields, protect the environment, and ensure the quality and safety of food and water
resources.
Understanding the Concept of IoT 29

2.11.7 Healthcare and Remote Monitoring

Among the sectors that stand to gain the most from IoT technology is healthcare. IoT
can assist healthcare professionals in lowering expenses, enhancing patient care, and
optimizing the use of scarce resources.
The ability to remotely monitor patients is one way that the IoT can benefit
healthcare providers. IoT-enabled gadgets, for instance, can be used to monitor vital
signs like blood pressure and pulse rate and transmit the collected data to medical
professionals for review. This can assist medical professionals in detecting health
issues early on, which can minimize the need for in-person visits and save expenses.

2.12 CHALLENGES AND FUTURE DIRECTIONS

By optimizing the flow of goods and services, businesses can provide their customers
with faster delivery and better service.

2.12.1 Interoperability and Integration

In the IoT, interoperability and integration are two key ideas that are vital to facilitating
effective data interchange and communication among diverse systems, applications,
and devices. In this chapter, we will examine the definitions of interoperability and
integration in the context of the IoT, as well as some possible advantages.
In the IoT, interoperability is the capacity of various systems and devices to
exchange data and communicate with one another.
The concept of interoperability is based on the fact that different IoT devices,
regardless of their manufacturers, should be able to talk to each other and share data
seamlessly. In other words, interoperability enables different devices, applications, and
systems within the IoT ecosystem to work together without requiring modifications
or additional efforts.
In contrast, integration in the context of IoT refers to the process of integrating
various systems and devices into a broader IoT ecosystem. Integration involves
bringing together various IoT elements, such as sensors, devices, applications, and
platforms, to create a cohesive and interconnected system. The goal of integration is
to enable seamless data exchange and interaction among these components.
By implementing interoperability and integration in IoT, various benefits can be
achieved:

1. Enhanced efficiency: Interoperability and integration in IoT allow different

devices and systems to communicate with each other seamlessly, resulting in
enhanced efficiency and reduced manual intervention.
2. Data integration: By integrating different devices and systems, IoT enables
seamless data exchange between different entities within the IoT ecosystem.
This facilitates data-driven decision-making, analytics, and actionable
insights.
30 IoT and Machine Learning for Smart Applications

3. Scalability and interoperability: Interoperability ensures that IoT devices can

integrate with other systems and devices, regardless of their manufacturers
or communication protocols. This promotes scalability and adaptability,
allowing the IoT ecosystem to grow and evolve without being limited by spe­
cific vendor lock-ins.
4. Reduced costs: Integration of IoT devices and systems can save costs by elim­
inating the need for separate systems or interfaces, streamlining operations,
and optimizing resource allocation.

For devices and systems inside the IoT ecosystem to communicate and exchange
data seamlessly, interoperability and integration are essential. By facilitating effi­
cient communication, integration promotes scalability, cost reduction, and improved
decision-making.

2.12.2 Scalability and Infrastructure

Infrastructure and scalability are essential components of the IoT. Scalability is the
capacity to manage growing numbers of devices and data without appreciable per­
formance reduction. Infrastructure is the supporting framework that enables connect­
ivity between devices and the cloud.
IoT systems need to be scalable in order to expand and change to meet shifting
needs. The system ought to be able to manage the rise in data and processing demands
as more devices are connected to the network without experiencing overload.
Infrastructure, on the other hand, refers to the physical and digital infrastructure
that supports IoT. This includes network connectivity, data centers, cloud infrastruc­
ture, and devices themselves.

2.13 CONCLUSION
We have discussed the key points, such as the ability to collect vast amounts of data,
the seamless connectivity of devices, and the automation of mundane tasks, all of
which contribute to the profound potential of IoT. This technology revolution has
the potential to improve people’s quality of life in addition to increasing operational
efficiency for corporations. IoT has enormous potential effects on both the business
and society.
Smart cities could lead to more sustainable and efficient living environments,
while industries can achieve unmatched levels of productivity and innovation. From
a societal perspective, the enhanced data and connectivity across devices can lead
to improved healthcare, smarter transportation systems, and more personalized con­
sumer experiences.
Looking forward to the future of IoT, we are on the cusp of a new era where every
object can be smart and connected. The possibilities are limitless, and the opportun­
ities for improvement and growth are boundless. While challenges such as security
and privacy remain, with the right regulations and ethical considerations in place, IoT
can lead humanity into a more interconnected and intelligent world.
Understanding the Concept of IoT 31

As we stand at the threshold of this exciting future, the importance of embracing

IoT cannot be overstated. It’s time for policymakers, business leaders, and individ­
uals alike to harness the power of IoT to shape a better tomorrow. So, let us take the
first step toward unlocking the full potential of the IoT by educating ourselves, advo­
cating for responsible use, and investing in this groundbreaking technology. Join the
movement and be a part of the IoT revolution!
 Unlocking the
3 Power of IoT
An In- Depth Exploration
Prakash Shanmurthy, Damodharan D.,
Anandhan Karunanithi, Pajany M., and
J. John Bennet

3.1 INTRODUCTION TO IOT

An Internet of Things (IoT) network consists of devices and entities that are
interconnected and can exchange information over the Internet. Anything from mun­
dane objects like domestic appliances, automobiles, wearable technologies, and even
industrial machinery to basic electrical devices like smartphones and laptops might
be included in these gadgets. The primitive objective of IoT is to make it feasible for
these entities to connect, communicate, and collaborate in a simple manner. This will
improve convenience, automation, and productivity in many facets of daily living [1].
The revolutionary concept that alludes to a very great extent network of interrelated
devices and items, all equipped with sensors, processors, and communication capabil­
ities are called IoT. These devices may be used to gather and distribute information
online without direct contact or a link between a person and a computer. Essentially,
IoT is about making these gadgets connect, communicate, and work together seam­
lessly, increasing productivity, efficiency, and convenience in many aspects of our
lives [2].
In IoT, “Things” can gamut from traditional electronic tools identical to
smartphones, computers, and suitable for wearing, to everyday objects like home
appliances, vehicles, industrial machines, and even sensors embedded in the environ­
ment. In the IoT, there are a number of connected devices that enable the production,
analysis, and utilization of real-time data.
In recent years, miniaturization, sensor technologies, cloud computing, and data
analytics have contributed to the growth of the IoT. With these elements, IoT can be
integrated into a various fields, including smart cities, smart homes, healthcare, trans­
portation, agriculture, and manufacturing [3].
The IoT is changing the way humans interact with the world by transforming inani­
mate objects into “smart” devices that are able to communicate, sense, and respond to
their environment and human interactions. An increasingly connected and data-driven
world will hold new opportunities for innovation and efficiency as the IoT continues
to evolve [4,5].

32 DOI: 10.1201/9781032623276-3
Unlocking the Power of IoT: An In-Depth Exploration 33

3.2 DEFINITION AND EVOLUTION OF IOT

The concept of connecting devices and enabling them to communicate has existed for
much longer than the term “Internet of Things” was coined. Several phases can be
identified in the evolution of IoT:

• Early Concepts and Radio Frequency Identification (RFID) (1990s-2000s): As

early as the late 1990s and early 2000s, the idea of connecting physical objects
to the Internet was explored. The initial focus was on RFID technology, which
made exploitation of radio signals to track and identify items. However, the
breadth and realization of these early conceptions were still incomplete [3].
• Proliferation of Internet and Connectivity (2000s): As Internet connectivity
became more widespread and affordable, the groundwork for IoT was laid.
In recent years, high-speed Internet and wireless communications, such as
Bluetooth and Wi-Fi, have significantly improved connectivity and communi­
cation between devices [6].
• Emergence of Smart Devices (2010s): Smart devices with sensors, processors,
and networking capabilities gave rise to IoT in the 2010s. Smartphones,
smartwatches, activity trackers, and smart home appliances were among these
gadgets. The IoT ecosystem is a result of these devices’ access to the Internet
and to one another.
• Machine-to-Machine (M2M) Communication (2010s): IoT has become a
major factor in the development of M2M communication, a method of trans­
ferring data directly between devices without human involvement. This
enabled automation and data exchange between devices, leading to improved
efficiency in various industries, such as manufacturing and logistics [1,3].
• Big data and Cloud computing (2010s): Big data and cloud computing tech­
nologies were necessary to handle the enormous amounts of data generated by
IoT devices. An affordable and scalable cloud service provides scalability and
storage. With big data analytics, it is now possible to gain valuable insights
from this data and to enable data-driven decisions to be made based on that
information [6,4].
• Intelligent IoT and Artificial Intelligence (2010s): The consolidation of arti­
ficial intelligence and machine learning with IoT devices brought about the
concept of intelligent IoT. Smart devices became capable of analyzing and
interpreting data on their own, making autonomous decisions, and adapting
to user preferences. This paved the way for applications like smart assistants,
predictive maintenance, and personalized user experiences [7,8].
• Industrial IoT (IIoT) and Smart Cities (2010s): IoT applications grew beyond
consumer gadgets as the technology developed. Manufacturing, energy, agri­
culture, and healthcare are just a few of the areas that have been transformed
by the IIoT idea. Similar to this, smart cities have gained a lot of attention.
They use IoT technology to improve urban infrastructure, streamline muni­
cipal services, and raise inhabitants’ quality of life in general [8].
• 5G Connectivity (2020s): The rollout of 5G technology in the 2020s has fur­
ther accelerated the evolution of IoT. With 5G, real-time communication is
34 IoT and Machine Learning for Smart Applications

enhanced, connected devices are enabled, data transfer speeds are faster, with
low latency, and connections are more plentiful [7,8,3].
• As technology continues to advance and more devices become connected,
a more connected and data-driven world is anticipated as a result of the IoT
ecosystem’s predicted exponential growth and new breakthroughs and uses
across several industries.

3.3 KEY COMPONENTS OF IOT

The IoT comprises various key aspects that operate collectively to facilitate seamless
connectivity, data exchange, and interaction between devices. These components
form the foundation of the IoT ecosystem. The essential factors of IoT include:

1. Device/Things: Devices, also known as “Things” in IoT, are physical objects

that are equipped with sensors, actuators, and communication modules. These
devices can be anything from intelligent phones, sensing devices, wearable
technologies, and domestic appliances to modern machinery, vehicles, and
environmental monitoring equipment.
2. Connectivity: The significant IoT-based system assists in the connectivity to
communicate and interchange data with each other. Various communication
technologies facilitate this, incorporating Bluetooth, cellular networks (3G,
4G, 5G), Wi-Fi, Z-wave, LoRaWAN, Zigbee, and NB-IoT [6,4,3].
3. Data Processing: A huge amount of data is produced by IoT from linked devices.
Data processing is gathering, storing, and analyzing this data in order to get an
insightful understanding and make defensible judgments. Data processing might
take place in centralized cloud servers or on the device itself (edge computing).
4. Cloud Computing: It provides scalable and secure storage and processing cap­
abilities for IoT data. Cloud platforms enable data aggregation, analysis, and
storage, making it accessible from anywhere and facilitating real-time data
exchange.
5. Data Analytics and Artificial Intelligence (AI): IoT devices spawned sub­
stantial volumes of data in a critical task for Statistical analytics and AI [3].
Automated decision-making, personalized suggestions, and predictive main­
tenance are made possible by AI algorithms’ ability to analyze data patterns,
spot abnormalities, and gain insightful knowledge [7,8,4].
6. User Interface: Interacting with IoT systems and devices is made possible by
the user interface component. This can be done using voice assistants, web
interfaces, mobile applications, or other graphical user interfaces, making it
simpler for consumers to manage and keep an eye on linked devices [6].
7. Security and Privacy: It is critical that IoT devices are secure, as they are sus­
ceptible to cyber threats and data breaches. Scrupulous protective assessment,
including encryption, endorsement, and acquire controls, are necessary toward
guarding IoT devices and the information they collect [9,2,8].
8. Standards and Protocols: IoT devices need to adhere to standard commu­
nication protocols to ensure interoperability and seamless data exchange.
 sanet.st

Unlocking the Power of loT: An In-Depth Exploration 35

Standardization efforts, such as Message Queuing Telemetry Transport

(MQTT), Constrained Application Protocol (CoAP), Hypertext Transfer
Protocol (HTTP), and Open Platform Communications Unified Architecture
(OPC-UA), help create a unified IoT ecosystem [3].
9. Edge Computing: In its place of transmitting all data to federal cloud servers,
edge computing administers data farther from where it originates (at the edge
of the network). Edge computing reduces latency, bandwidth usage, and
optimizes real-time adaptability, rendering it perfect for IoT applications that
require quick response times [7,8].
10. Firmware and Over-the-Air (OTA) Renews: Firmware upgrades for IoT
devices are frequently needed to incorporate new functionalities, fix bugs, or
enhance security. OTA updates allow devices to be updated remotely without
physically accessing them.

These essential IoT elements work together to provide a connected ecosystem

that enables devices to interact, collect and analyze data, and offer intelligent and
automated services. As a result, efficiency, convenience, and innovation are ultim­
ately increased across a variety of sectors and disciplines [5].

3.4 APPLICATION AND USE CASE OF IOT

There have been a number of applications of IoT in a variety of domains, including
the following:

1. Smart Homes: Home automation is made possible by IoT, allowing users

to operate appliances like smart thermostats, lighting controls, and security
cameras remotely using their smartphones or voice commands.
2. Healthcare: IoT devices such as wearables and remote monitoring systems help
in tracking health parameters and providing real-time health data to healthcare
professionals, leading to better diagnosis and personalized treatment [7,6].
3. IIoT: IoT is utilized in manufacturing and industry to enhance efficiency and
decrease downtime through proactive maintenance, process optimization, and
asset tracking [8,5].
4. Smart cities: The IoT is now being used in a variety of ways in order to
improve urban infrastructure, control traffic, optimize energy consumption,
and advance overall city services [2,7].
5. Agriculture: IoT-enabled sensors in agriculture support optimizing irrigation
and increasing yields, keeping an eye on crop fitness, climate patterns, and
soil specifications [7].

3.5 CHALLENGES AND OPPORTUNITIES IN IOT

3.5.1 Challenges
Huge prospects for modernization and connection are provided by IoT. Nonetheless,
it also poses a number of difficulties that must be resolved to guarantee its
36 IoT and Machine Learning for Smart Applications

implementation’s success and security. IoT is currently facing a number of challenges,

some of which are listed below:

• Privacy and Security: As IoT devices often transmit and gather sensitive infor­
mation, they are vulnerable to hacking and privacy violations. Unauthorized
data retrieval or manipulation of connected devices might result from inad­
equate security measures and inadequate encryption [9,2].
• Interoperability: The IoT ecosystem consists of numerous devices and
platforms from different manufacturers. The development of cohesive IoT
solutions can be hindered by the difficulty of ensuring seamless communica­
tion and compatibility between these devices.
• Scalability: IoT network’s infrastructure management and scalability become
a rise in difficulty as additional linked devices are added. As the number of
devices increases, it is more difficult to ensure that the system remains effi­
cient and reliable.
• Data Management and Analytics: IoT creates a tons of data from many
sources. It might be difficult to adequately organize and interpret this data in
order to produce practical insights; this requires strong data analytics capabil­
ities [7,2,4,3].
• Power Consumption: The majority of IoT devices are battery-powered, and
optimizing power consumption is essential in order to extend their lifespan
and reduce the frequency with which they need to be replaced or recharged.
• Connectivity Issues: Internet access must be dependable and consistent for
IoT. Data transfer and device control may be difficult in places with patchy
network coverage or inconsistent connectivity.
• Standards and Regulations: The lack of universal standards and regulations
in the IoT industry can create challenges in terms of data security, interoper­
ability, and compliance with legal requirements [3].
• Cost: IoT infrastructure and device deployment can be expensive at first, espe­
cially for large-scale applications. Widespread adoption may be hampered by
this expense, particularly in certain sectors or geographical areas [4].
• Privacy Concerns: IoT devices are embedded in various aspects of daily life,
raising concerns about constant data collection and potential invasion of
privacy [9,2,3,10].
• Environmental Impact: When IoT devices proliferate quickly and are not
properly recycled or disposed of this can result in electronic waste and envir­
onmental problems [7].
• Reliability and Resilience: IoT devices and systems need to be highly reliable
and resilient, notably in mission-critical settings like medical care, transporta­
tion, and business operations where failure can have serious repercussions [3].

Collaboration among a variety of stakeholders, including producers, legislators,

researchers, and consumers, is necessary to address these issues. Strong security
measures, adherence to privacy laws, and the development of open standards can all
help to address some of the major IoT concerns and unlock the full potential of the
technology across a variety of industries [7].
Unlocking the Power of IoT: An In-Depth Exploration 37

3.5.2 Opportunities
An array of opportunities are offered by the IoT across a wide range of industries and
fields. As technology develops and linked gadgets keep multiplying the opportunities
in IoT become increasingly compelling. Some of the key opportunities in IoT include:

Enhanced Efficiency: The IoT enables automation and real-time data analysis and
improves the efficiency of processes and operations. Some of the efficiency-driven
opportunities in IoT include optimizing operations, eliminating waste, and opti­
mizing resource consumption.

Improved Decision Making: Through its network of linked devices, IoT produces col­
ossal volumes of data. Businesses and groups may be able to maximize their strategies
and make decisions based on data by examining this data and gaining insightful insights.

Personalized and Contextual Experiences: By gathering and analyzing information

provided by connected gadgets, IoT can enable customized services for consumers.
This makes it possible to provide customized services like custom suggestions,
automated smart homes, and targeted advertising [2].

Healthcare Advancements: Remote tracking of patients, wearable medical tech­

nology, and telemedicine services are all ways that IoT is revolutionizing the healthcare
sector. It enables proactive and individualized treatment, improving patient outcomes,
and lowering healthcare expenditures [7,6].

Smart Cities: IoT is revolutionizing urban living by making cities smarter, more
efficient, and sustainable. Smart cities cover a wide range of topics, including smart
mobility, energy management, waste management, and public safety [2,7].

Industry 4.0 and Smart Manufacturing: IoT is driving the fourth industrial revo­
lution, where connected machines, sensors, and data analytics enable smart manu­
facturing and intelligent supply chain management. As a result, production rises,
downtime decreases, and the quality of the final product improves [8,5].

Agricultural Innovation: IoT-based precision agriculture is transforming farming

practices, enabling farmers to monitor crops, optimize irrigation, and improve
resource utilization. This promotes secure access to food and sustainable agriculture.

Connectivity and Communication: IoT makes it possible for people and things to
communicate and interact seamlessly. This enables real-time collaboration, remote
monitoring, and improved communication channels.

Environmental Monitoring and Conservation: IoT devices are capable of detecting

animals and monitoring environmental factors, such as the quality of the air and water.
Using this knowledge, the ecosystem might be conserved, and the impacts of climate
change may be mitigated [7].
38 IoT and Machine Learning for Smart Applications

Safety and Security: IoT technologies provide improved security and safety
measures, including surveillance, asset monitoring, and smart home security systems.
These advancements help in preventing accidents, reducing crime rates, and protecting
valuable assets [2].

Smart Energy Management: Energy use in homes, buildings, and enterprises may
be better controlled and monitored owing to the IoT. This leads to optimized energy
usage, cost savings, and reduced carbon emissions [6].

Innovation and Entrepreneurship: The growth of IoT presents numerous oppor­

tunities for entrepreneurs and startups to develop innovative solutions and disrupt
existing industries.
IoT delivers a wide range of options for organizations, individuals, and commu­
nities. Embracing IoT technologies and leveraging their potential can lead to trans­
formative changes in how we live, work, and interact with the world around us.

3.6 ARCHITECTURAL LAYERS OF IOT

3.6.1 Introduction
IoT architecture is the combination of sensors, software, communication protocols,
vehicles, intelligent devices, buildings, physical objects, cloud services, users,
developers, business layers, actuators, etc. Data management systems gather data for
storage, analysis, processing, and exchange. All these network physical components,
configuration, practical organization measures, and working values.

3.6.2 Real-World Components of IOT

The practical components of IoT encompasses of several layers. They are:

1. Application and service

2. Detection
3. Announcement
4. Cloud
5. Administration

3.6.2.1 Applications and Service

This layer provides a variety of services, which include the collection of data, ana­
lyzing the data, visualizing the data, and providing security to the data. All these
depend on the use cases and features of the end users.

3.6.2.2 Detection
An important function of the sensing layer is to detect fluctuations in the physical
form of the practical components during actual stretching. The main component of
this layer is sensors, which are responsible for recognizing and positioning the intel­
lectual objects, gathering the information, and transporting the information to the
Unlocking the Power of IoT: An In-Depth Exploration 39

cloud environment for processing and storing the information. In this layer, we can
find an actuator, which is a mechanical device such as a switch that is used to execute
the chosen device.

3.6.2.3 Announcement
The communication layer has certain devices like gateways, routers, and switches,
which are connected to devices that cannot link to the cloud services directly. This
layer is accountable for interactions or transfer of information to the other IoT layers.
The data which are composed by the sensing layer will be shifted to the cloud service
and the application layers.

3.6.2.4 Cloud
Universally, the function of the cloud is to use a data center as a fundamental server
to develop the generated information by the edge device. Information composed
from sensors and devices are consumed into the cloud layer. The purpose of cloud
is to store data, processing of data, and finally analyze the information that is
available. In different circumstances we can say this cloud layer as the IoT system
processing unit.

3.6.2.5 Administration
The responsibility of the administration layer is to operate and monitor all the
other layers with the help of cloud administration tools, which are generally
employed [11].

3.6.3 IOT Architecture

Designing a flexible architectural outline is vital for IoT module at networking with
huge amount of diversity over the Internet. The major concerns to be considered
while developing the architecture of an IoT network are

• Privacy and security

• Data storage and Quality of Service (QoS)
• Scalability
• Reliability

The absence of a solid architectural pattern has led to numerous initiatives. However,
no standard model has been adopted due to the wide range of applications and outline
plans with diverse variables and design patterns. In order to scale the computation of
the current classifications, the standard model has not been implemented.
Despite this, the most commonly used IoT architectures are:

1. Three-layer Architecture.
2. Middleware-based Architecture.
3. Service-Oriented Architecture.
40 IoT and Machine Learning for Smart Applications

3.6.4 Three-Layer Architecture

IoT research began with the development of the three-layer architecture, which has
been referred to as the most fundamental architecture. The architecture is divided into
three layers, such as

1. The application layer describes how the computer software contributes to the
exact provision of a service.
2. In the network layer, information is broadcast to the upper layers.
3. The Perception Layer describes the physical surface of an object and clusters
all its features together.

3.6.4.1 Application Layer

The IoT Architecture begins with the application surface. It lies in between the user
program and the application program. A network layer administrator is respon­
sible for retrieving information from a network layer and utilizing information that
has been composed for explicit purposes. The information will be used for ware­
housing, accumulating, filtering and managing, databases, etc. As an outcome of
this, the information will be accessible to applications of IoT such as intelligent
cars, wearables, intelligent cities, intelligent transportation, intelligent agricul­
ture, healthcare systems, etc. Furthermore, this layer analyzes the data that has
already been received in the application layer and stored there to forecast the future
condition of physical objects [6]. The most common software technologies that
widely use this massive amount of information by the appliances are edge and
cloud computing. The arrangement of the information to be processed can be any
type as below:

• Small and binary information cannot be read by human beings;

• Larger text information is readable by human beings [12].

3.6.4.2 Network Layer

This network layer may be referred to as a communication layer, which is merged
with the middle layer of the IoT, which is considered to be the fundamental com­
ponent of the IoT ecosystem. Providing routing frequencies for information trans­
mission should be the responsibility of the communication layer. The network layer
is responsible for providing connectivity, transferring messages between various
devices, cloud service providers, and countless servers. This layer acts as a boundary
for perception and application layers, which helps IoT devices, applications, and
servers to receive the information. Network layer consists of different communication
standards, such as Bluetooth, Wi-Fi, and IoT objects. These devices will be useful in
critical circumstances with the help of wireless protocols, which play an important
role in this layer because they require less material and human effort. In addition to
routing capabilities and unique addressing, IoT promises easy integration of various
devices into the same device [13].
Unlocking the Power of IoT: An In-Depth Exploration 41

3.6.4.3 Perception Layer

The persistence of this layer is achieved by collecting certain information about the
objects, and it has to identify the objects that are unique in the IoT ecosystems. This
layer consists of sensors that transmit and receive information from other layers of the
atmosphere, as well as processing the information in the upper layers [14]. The per­
ception layer embraces objects that can communicate with peripheral devices which
can sense with some computing capabilities, in general terms, we can say that as
smart technologies.
IoT uses these smart technologies as the fundamental blocks that can be used for
common purposes (smart watches, smart TVs, etc.). Smart objects were equipped
with some general properties like sensing and actuation, addressability, communi­
cation, embedded data processing, identification, localization, and user interface.
All these properties use a common factor called sensors to collect and communi­
cate the information with other layers. Examples: Arduino, Raspberry Pi, Beaglebone
Black, etc.

3.6.5 Security and Confidentiality

Every system will have confidentiality and security issues that should be addressed
to protect the IoT applications. There are some security issues associated with each
layer of the IoT architecture, which should be taken into consideration at the very
beginning of the design process. We will discuss some security issues in the IoT
architecture at each layer.

3.6.6 Security Issues in the Application Layer

As per the necessities of the applications, this layer requires different security
standards. Some listed privacy and security issues of this layer are:

1. Application-specific liabilities: Some liabilities may be left which was

unknown to the users, which also creates a security threat to the user’s
information.
2. Information administration: There is an increase in complexity of the system
as the collection of data increases, leading to a demand for creating resources
and requiring complex algorithms to administer the information, which may
lead to information loss.
3. User information confidentiality: At each communication, users’ private
information should be protected from vulnerable attacks. Sometimes, the
technique which is used to process the information may lead to vulnerable
activities, which may lead to huge loss of information.
4. Mutual confirmation and node identification: There are various levels of
access privileges required by each application due to the presence of various
users. It is essential to apply proper authentication procedures in order to pre­
vent unauthorized access.
42 IoT and Machine Learning for Smart Applications

3.7 COMMUNICATION PROTOCOL FOR IOT

An important technology of the IoT is ZigBee, which enables users to benefit from
smart applications and services. However, IPv6 transports an extreme of 1280 bytes of
packets, ZigBee empowered instruments cannot have a grip on it. By using a gateway,
these IPv6 packets will be transmitted over ZigBee networks and may pass through
ZigBee Coordinator [15]. The main function of this ZigBee Coordinator and gateway
is to discover the process which are close to the network. This creates a complexity
problem for the ZigBee Coordinator. In addition to these routing structures, data
forwarding issues, and header size problems are also there for ZigBee devices, which
needs to get addressed [3,10-19]. As a solution to these issues for all loT-enabled
ZigBee devices, we can propose a peer-to-peer communication protocol hosted using
the IPv6 over Low-Power Wireless Personal Area Networks (6LoWPAN) protocol
over the Internet, which reduces the burdens on the ZigBee Coordinator.
The gateway used here will act as a protocol translator for passing the messages
through with hosted Internet and with ZigBee node. It decodes the received messages
to ZigBee frames, which originate as IPv4 packets travel through IPv6 packets, which
demands hardware and software support, including the ZigBee devices that suffer
from a header size problem [17].
Access to the whole network will be with the coordinator node. Once the coordin­
ator node fails to function, it needs to be re-booted, which is not likely to execute, and
it leads to the failure of the whole network in a single point of failure [15].

3.7.1 Thought-Provoking Problems with ZigBee-Enabled Devices

1. The peer-to-peer transmission with Internet hosts and ZigBee nodes.
2. A coordinator node controls the Zigbee network.
3. Hardware and software need fresh entry point.
4. Header dimensions problem.
5. Single point of failure.

To overcome these issues in an IoT-based network, we can use the 6LoWPAN com­
munication protocol, which attaches a group of Zigbee devices with an Internet
protocol-based setup that affords peer-peer announcement to direct ZigBee-enabled
IoT networks with the IPv6 [18].
Three primary services will be offered when the protocol has an enquiry from the
user, they are:

i. Disintegration and refabrication used.

ii. Title firmness.
iii. Link-layer forwarding.

3.7.2 ZigBee Communication Protocol

ZigBee protocol enables transmission of information with the edge nodes and the
claim outlines, which supports low power protocol and low cost. This protocol
Unlocking the Power of IoT: An In-Depth Exploration 43

FIGURE 3.1 ZigBee protocol layer.

consists of five different layers that perform different functionalities to provide the
proper communication between the nodes which are all connected [19]. The ZigBee
Protocol layer is shown in Figure 3.1.
The five building blocks used to develop the ZigBee communication protocol are:

1. Application or Request Block

2. Network or Mesh Block
3. 6LoWPAN Block
4. Media Access Control (MAC) Block
5. Physical Layer (PHY) Block

3.7.3 6LOWPAN Block

The block is also called the 6LoWPAN procedure block, which is an amendment
block and provides amenities to the Internet blocks used to collect the information
which are available at the end nodes via MAC and PHY block, consolidates the
collected information and handovers it over to the next upper layer. ND is used to find
the neighbor nodes after that create the default communicating routes with the nodes
and then the transmission starts [15].
The services provided by this protocol are shrinking header files, disintegration
and refabrication, programmed formation, and neighbor detection in the upper block.
Using shared information, the shrinking of header files service reduces the size of
User Datagram Protocol and IPv6 header fields, which can be misplaced when header
fields are removed from associated blocks [19].

3.7.4 MAC Block

The block offers three mechanisms: (i) Guided mode, (ii) Unguided mode, and (iii)
General Switching Transceiver (GST) distribution. To accomplish non-breakable
communication, communicating devices depend on the salutation, which can be
done without signals. It will be helpful to save a huge amount of power as emitting
44 IoT and Machine Learning for Smart Applications

signals at each and every triggering of the device will loss huge amount of power
[18]. ZigBee communication networks use guided transmission frequency access
to expand their potential and extend their snooze intervals. The routers in ZigBee
occasionally release signals at all signaling points to broadcast their presence. The
communicated signal edges will work as a register, which varies between 15 Ms and
252 s. Among each signal break, the nodes can snooze, dropping duty routines, which
increases battery life and lowers latency.

3.7.5 PHY Layer

This block has the responsibility of transmitting and reception of datagrams over
a physical medium, which facilitates initiating and defusing a radio transceiver,
energy recognition, joining eminence signal, frequency range, and the evaluation of
the channel to the upper block. This block functions on a dual-band, which ranges
from 2.4 gigahertz to 868-916 megahertz based on Direct Sequence Spread Spectrum
[19]. Lesser propagation loss will occur when low transfer rate occurs at low band
with larger range of functions. Apart from that a better frequency outcome in quicker
amount of data, delay in time, and reduces burden [15].

3.8 IOT HARDWARE AND SENSORS

3.8.1 IOT Hardware

In the IoT, a hardware device is a physical device and mechanism designed to collect
information, connect to devices, and communicate with devices connected via the
Internet [20]. IoT hardware comes in various forms, sizes, and complexity levels,
depending on the use case and application domain. As a result of IoT technologies,
a growing number of sophisticated and diverse IoT hardware solutions have been
developed [21].
The commonly used IoT hardware interfaces:

1. Camera and Imaging Modules: IoT devices are unified with cameras for
object detection surveillance [22].
2. Global Positioning System (GPS): IoT devices may include different
positioning modules, which provide information on locating objects over the
devices through the Internet [23].
3. Edge Computing Devices: IoT applications are commonly used in edge
computing devices to process the data locally instead of transferring all data
to the cloud, which reduces latency and saves bandwidth [24].
4. Display Interfaces: Most IoT devices use display interfaces to communi­
cate with the outsiders, such as Liquid Crystal Display (LCD), Light Emitting
Diode (LED), and touchscreen displays to interact with the users [20].
5. Gateways: To maintain communication between IoT devices and other
existing networks a gateway is a common technique used to collect informa­
tion from multiple devices, develop the information and then send the infor­
mation to other networks for further usage and storage.
Unlocking the Power of IoT: An In-Depth Exploration 45

6. Connectivity Modules: To exchange the processed information, IoT devices

need to get connected to the Internet. Such communications can be established
through cellular networks, Wi-Fi medium, Bluetooth, etc. [21].
7. Actuators: Based on the instructions received from the microcontrollers,
allow the IoT devices to interact with the physical world.
8. Sensors: Sensors are the components that collect real-world information from
the physical environment and detect the physical parameters such as tempera­
ture, humidity, light, motion, proximity, pressure, gas levels, etc., for each
type of parameters different types of sensors are used based on the specific
application [20].
9. Microcontrollers and Microprocessors: Microcontrollers and
microprocessors are the Central Processing Unit (CPU) of IoT devices used
to handle data processing, decision-making, and control of connected sensors
and actuators.

Examples include Arduino, Raspberry Pi, etc. [24].

3.8.2 IOT Sensors

IoT sensors are a kind of hardware device which is used to collect information from
outside the physical world or devices and translate that information into digital infor­
mation which can be processed and communicated over the Internet [20].
These sensors are fundamental functioning unit of IoT systems helps in enabling
various applications across different domains.

3.8.3 Common Types of IOT Sensors

Temperature sensors: which allows the users to scale the temperature, which can be used
in applications such as climate control, weather monitoring, and industrial processes.

Humidity sensors: which helps in scaling the level of moisture in the air and are used
in applications like Heating, Ventilation, and Air Conditioning (HVAC) systems, agri­
culture, and weather stations [21].

Light sensors (photodetectors): Detect the light levels in the physical world and
are utilized by applications like automatic lighting control, display brightness adjust­
ment, and outdoor lighting.

Motion sensors: Detect the movement or change in position and are used in security
systems, smart lighting, and gaming applications [23].

Accelerometers: Scale the acceleration forces and are commonly used in fitness
trackers, smartwatches, and motion-based applications.

Pressure sensors: Scale the atmospheric or fluid pressure and find applications in
weather forecasting, altitude measurement, and industrial monitoring [24].
46 IoT and Machine Learning for Smart Applications

Gas sensors: Detect the existence of several gases in the atmosphere and are useful
in quality monitoring and safety in industrial settings [23].

Sound sensors (microphones): Captures audio signals and is used in voice recogni­
tion noise monitoring.

Ultrasonic sensors: Use sound waves to detect the remoteness of an entity and are
widely used in robotics, parking assistance systems, and industrial automation [21].

Image sensors (cameras): Capture the visual data and utilize that information in
applications like surveillance systems, facial recognition, and object detection.

Stress sensors: Measure the stress and are used in health monitoring and industrial
applications.
These are a few examples of the wide variety of IoT sensors available. Each type of
sensor serves specific purposes in different IoT applications, making them valuable
components in building smart and interconnected systems [25].

3.9 IOT WITH CLOUD COMPUTING (CC)

3.9.1 Introduction
IoT and CC are the two best examples that have gained attractions from the academics.
Cloud computing provides services to users as an on-demand service that helps them
avail the services from anywhere [26]. On the other hand, IoT uses hardware and
sensors to provide service to the users as they smell changes in the physical world,
and transmit information to the users who access it. We will discuss how both tech­
nologies work together to improve the betterment of the existing work [27].

3.9.2 Fusion of IOT and CC

IoT and CC have grown independently and created most of the impact in the technical
field, but they are still reaching different levels. If the two technologies are combined,
we can find huge benefits in the technical field.
Cloud computing is creating almost infinite possibilities in generating the huge
amount of hardware and software like infrastructure, storage, and applications as a
virtualized environment, which creates an advantage to IoT. When these cloud ser­
vices and IoT devices come together, it creates an impact which can be considered as
an evaluation [26].
As cloud computing is getting benefitted from IoT, IoT is also getting benefitted,
such as endless volume and properties can be provided for the information which are
collected by the IoT devices [28]. Cloud computing will significantly benefit from the
widespread and advanced implementation of IoT, as it facilitates seamless commu­
nication between various devices in the physical environment [26]. By enabling this
connectivity, cloud computing can provide innovative solutions and services tailored
Unlocking the Power of IoT: An In-Depth Exploration 47

to a wide array of real-life scenarios, acting as a crucial intermediate block among

things (devices) and requests.
Additionally, the connection between CC and the IoT creates matchless opportun­
ities to collect, integrate, sharing data with others [29]. To ensure continuous avail­
ability and security, information from the cloud should be well-matched with a typical
API, which permits it to benefit from healthy protective measures that are accessible
from any location. Additionally, the dealing of IoT nodes doesn’t necessarily need
to occur within the devices themselves, as the data can be efficiently integrated and
transferred to more capable storage nodes [30].

3.9.3 Services Provided

Merging of IoT and cloud services helps in developing functional services like:

• Sensing as a Service (SaaS)

• Database as a Service (DBaaS)
• Sensing and Actuation as a Service (SAaaS)
• A sensor as a Service (SenaaS)

SaaS: It is anticipated to build on top of the foundation of IoT structure and ser­
vices to help in reaching the highest level of development and implementation. Using
the SaaS cloud model, sensing servers can effectively manage and process sensing
requests originating from various locations.

SAaaS: Cloud-based deployment of automated control logic is possible. For the

deployment of SAaaS, sensors and actuators must be provided as reliable and well-
defined services, bridging the gap regarding sensor networks and devices or compat­
ible principles.

SenaaS: facilitates the all-encompassing management of distant sensors, allowing for

their seamless and widespread control and monitoring.

DBaaS: It is a cloud-based, self-service method that enables users to deploy the

collection of information and their applications directly to the intended recipients
without requiring assistance from the processing department [26].

3.9.4 Benefits
Prior studies have demonstrated the broader benefits of combining IoT with cloud
computing, encompassing advantages such as enhanced storage capacity, increased
computational resources, and access to novel capabilities like advanced computing
technologies and protocols. Additionally, it offers improved scalability, accuracy, and
accessibility, which are typically challenging to obtain otherwise [31]. These benefits
will prompt companies and individuals to contemplate the ways they can leverage the
enabled technologies. This could aid organizations in addressing challenges related
48 IoT and Machine Learning for Smart Applications

to cloud computing, such as limited financial resources and infrastructure, while also
assisting them in making sense of their IoT information.

3.9.5 Demerits
The various challenges and complexities are.

i. Software systems in smart gadgets.

ii. Stabilization.
iii. Establishing interoperability.

Concerns and challenges related to IoT cloud often led organizations to hesitate or
be reluctant to adopt cloud technologies. Nevertheless, despite potential challenges
and issues, it’s essential to recognize that cloud computing also brings additional
benefits [26].

3.10 EDGE COMPUTING AND IOT

3.10.1 Introduction
Using edge computing, advanced computing capabilities are brought closer to
IoT nodes, facilitating the support of complex applications in the future IoT. By
deploying powerful computing resources at the edge that are closer to users, edge
computing offers real-time computing services [32]. Edge computing is used to
decentralize computing power, moving it away from centralized cloud servers, and
instead, placing it in edge nodes located near the end-users [33]. This approach
yields two significant enhancements over traditional cloud computing. First, edge
nodes can perform data preprocessing on a large scale before transmitting the data
to central cloud servers.
The result is a reduction in the amount of data that needs to be transferred,
resulting in an improved data handling process. Second, edge computing optimizes
cloud resources by endowing edge nodes with their own computing capabil­
ities [34].

3.10.2 Edge Computing and IOT

An edge server is positioned at the network edge and connected directly to the
access point (AP) of an Edge IoT network. There are numerous IoT nodes scattered
throughout the area surrounding the access point. The edge server serves as a local
computing hub, enabling real-time processing and analysis for the IoT devices and
applications within its proximity [35].
These IoT nodes encompass various smart devices, sensors, actuators, or other
objects equipped with embedded computing capabilities, forming an integral part
of the IoT ecosystem. As these IoT nodes operate and generate data, the edge server
plays a vital role in handling, aggregating, and responding to the information they
produce [32].
Unlocking the Power of IoT: An In-Depth Exploration 49

3.10.3 Edge Computing Implementation

During the implementation of edge computing architecture, numerous research
endeavors have been undertaken, resulting in several dominant models. These
prevailing models have emerged as the most influential and widely adopted approaches
in the field of edge computing [33].

1. Hierarchical model
2. Software-defined model

3.10.4 Integration of IOT and Edge Computing

Even though they operate independently, the integration of an edge computing plat­
form can significantly assist IoT in resolving critical challenges and enhancing per­
formance. By combining edge computing capabilities with IoT systems, various
issues related to latency, bandwidth efficiency, data processing, and response times
can be effectively addressed. Figure 3.2 shows the Layer Architecture of IoT-based
Edge-Computing.
The likelihood of combining IoT with edge computing is with a particular
emphasis on transmission, storage, and computation aspects. By narrowing our focus
on these characteristics, we aim to demonstrate how edge computing can enhance the
overall performance of IoT systems. Through this integration, IoT data transmission
becomes more efficient, data storage and processing become faster and more respon­
sive, leading to improved IoT performance and user experience.
Universally, IoT derives significant benefits from Edge and Cloud computing due
to distinctive features offered by these dual constructions. Edge computing provides
the advantage of high computational capacity and low-latency processing, thanks to
its deployment closer to the IoT devices and sensors [36].
On the other hand, cloud computing offers extensive storage capabilities, allowing
IoT data to be securely stored and easily accessible from anywhere. By harnessing the
strengths of both edge and cloud computing, IoT systems can achieve a more robust
and efficient ecosystem that optimizes both computational power and data storage.
Despite having inadequate computational dimensions and storing capacity, when
compared to cloud computing, edge computing still offers distinct benefits for IoT.
Additionally, edge computing excels in delivering fast response times, enabling real­
time meeting out and study of data directly by the side of the network’s edge. This
capability ensures that IoT applications can promptly respond to events and deliver
timely actions [33].

3.10.5 Advantages
The integration of IoT with edge computing offers several advantages. By com­
bining these two technologies, IoT systems can benefit from the reduction of latency,
improved actual processing, improved information confidentiality and safety,
optimized packet utilization, increased scalability, and more efficient use of com­
putational resources [35]. As a result of edge computing, data can be processed and
50 IoT and Machine Learning for Smart Applications

FIGURE 3.2 Layer architecture of IoT based edge-computing.

Unlocking the Power of IoT: An In-Depth Exploration 51

analyzed in a localized manner, allowing for faster response times and better overall
performance [34].

1. Transmission
a. Latency/delay, bandwidth, energy, overhead
2. Storage
a. Storage balancing
b. Recovery policy
i. Availability
ii. Data replication
3. Computation
a. Computation offloading
i. Local
ii. Edge/cloudlet
iii. Cloud
b. Pricing policy
i. Single service provider
ii. Multiple service providers
c. Priority

3.10.6 Disadvantages
The integration of edge computing to support IoT brings forth numerous advantages.
However, alongside these benefits, there are also challenges specific to edge
computing-based IoT that need to be addressed [36]. These challenges may arise
due to factors such as network connectivity issues, security concerns, scalability
limitations, and the complexity of managing distributed computing resources at the
edge. Edge Computing can significantly enhance IoT applications and services by
identifying and overcoming these challenges [33].

1. System integration
2. Resource management
3. Security and privacy
4. Advanced communication
5. Smart system support

3.11 IOT DATA MANAGEMENT

3.11.1 Introduction
As a paradigm, the IoT has emerged in recent years, enabling billions of devices
to communicate and generate vast quantities of data. Data management and ana­
lysis are essential for facilitating decision-making processes and gaining meaningful
insights from them. As IoT-generated data is large, fast, varied, and verifiable, the
study explores data management strategies tailored specifically for IoT environments.
Sensors, actuators, and connected devices are all part of IoT data management, which
52 IoT and Machine Learning for Smart Applications

involves gathering and integrating heterogeneous data from a variety of sources [37].
A network of interconnected devices, including phones, tablets, and wearables, the
IoT communicates and exchanges data through the Internet. IoT has the potential to
transform a variety of industries and aspects of our daily lives, in addition to enab­
ling automation, improving efficiency, and facilitating intelligent decision-making.
Data from sensors is used by edge gateway devices to make decisions through data
analysis. Data must be collected in real time, organized, and communicated between
these high-performance systems in real time. Data flow can be monitored by end
users using web-based dashboards running on edge devices. When searching a large
database containing a lot of data, graphs over arbitrary time periods can be used
to locate records. It involves a variety of technologies and techniques, such as data
warehouses, data lakes, and machine learning. IoT data management can be used for
a variety of purposes, such as improving operational efficiency. IoT data can be used
to identify inefficiencies in operations and make changes to improve efficiency. For
example, IoT data can be used to track the performance of machines in a factory and
identify machines that are not operating at peak efficiency. This information can be
used to make changes to the production process, which can lead to increased prod­
uctivity and decreased costs. Improving customer service: Customers’ behavior and
preferences can be tracked using IoT data. For example, IoT data can be used to
track the browsing history of website visitors and recommend products that they are
likely to find interesting. The result can be an increase in sales and a greater level of
customer satisfaction for businesses. IoT data can be used to identify new market
opportunities and to develop new products and services. For example, a supply chain
can be tracked using data collected from IoT devices. In order to improve efficiency
and lower costs, new logistics solutions can be developed based on this information.
Automating tasks and improving efficiency can be achieved. An example of the use of
IoT data is the tracking of inventory levels in a warehouse and the automatic ordering
of additional inventory when levels decline [38].

3.11.2 Data Collection and Acquisition

Collecting and acquiring IoT data involves grouping data from IoT devices and
making it available for analysis. Any IoT research must begin with this step, as it
enables businesses to collect the data they need to make informed decisions and
improve their operations. Decide what data will be collected? What data do you need
to collect? This will depend on the specific application of your IoT system. Using
IoT sensors, for example, to monitor the temperature of a manufacturing plant, you
will need to collect temperature data at various points throughout the facility. Choose
the right sensors. A wide variety of sensors are available, each with its strengths and
weaknesses. You need to choose sensors that are accurate, reliable, and that can collect
the data you need. Connect the sensors to the Internet. Once you’ve chosen the right
sensors, you need to connect them to the Internet. In order to accomplish this, several
technologies can be used, including Wi-Fi, Bluetooth, and cellular networks. A secure
location must be selected to store the collected data once it has been collected. All
can store the data on a local server, in the cloud, or in a combination of both. Many
sensors are used to collect data from different environments. This will help us to
Unlocking the Power of IoT: An In-Depth Exploration 53

collect data on a variety of factors, such as temperature, humidity, vibration, and

motion. Store the data securely such as the data we collect should be stored securely
in a location that is accessible to authorized users only. Make the most of your data
by using data analytics tools. We should use data analytics tools to get the most out of
our data. With the help of these tools, you are able to identify patterns and trends in
your data, and then make informed decisions based on this information.
IoT data collection and acquisition: Increasingly, IoT devices are collecting large
volumes of data. According to estimates, there will be 44.2 zettabytes of IoT data
globally by 2020. By 2025, this number is expected to reach 181 zettabytes. The var­
iety of data collected by IoT devices is also increasing. In addition to traditional data
types such as temperature, humidity, and vibration, IoT devices are now collecting
data on many factors, namely location, energy consumption, and customer behavior.
The speed at which data is collected by IoT devices is also increasing. In the past, IoT
devices would collect data at intervals of minutes, hours, or even days. Today, many
IoT devices are collecting data in real-time. There are some emerging trends in IoT
data collection and acquisition such as Edge Computing (EC), Cloud Computing (CC),
Artificial Intelligence (AI), and Data Analytics (DA). There are various challenges
faced during the data collection in IoT, such as compatibility with existing systems,
huge amounts of data need to sort, real-time data need to stream. The recent research
aims to bring a novel framework for IoT devices and a lightweight design to interact
with various sensors in real-time data collection. For data transmission over the net­
work a common protocol has been used such as HTTP, CoAP, and MQTT in the cloud
storage [39].

3.11.3 Data Storage and Processing in IOT

The different use cases need to store the data in different formats; however, each
device has the capacity to store data [39]. The worldwide is producing terabytes of
information with various sources produced using IoT. Every generated data is essen­
tial for extracting useful information [40].

3.11.3.1 Cloud Edge Collaborative Storage (CECS)

The cloud server and IoT device are interconnected using CECS. CECS is clearly
founded on the fact that cloud servers are wealthy in terms of capacity and com­
puting assets, edge servers are situated near IoT devices, and IoT devices usually have
limited assets [41]. Data will first be collected from IoT devices and sent to nearby
edges, where they will be processed in real time, returned results, and stored on cloud
servers. The user can ultimately share the IoT data expected on the cloud server by
creating relevant search queries.

3.11.3.2 P2P File System Based on Blockchain Technology

The framework for verifying the integrity of large-scale IoT data using blockchain
technology is shown in Figure 3.3. It is recommended that IoT devices sign blockchain
transactions or that other nodes verify the device’s signature. IoT devices’ flaws can
be fixed by using edge servers. Several IoT devices near the edge server constantly
54 IoT and Machine Learning for Smart Applications

FIGURE 3.3 Internet of things (IoT) general solution architecture and storage.

collect data [42]. In this case, it determines which addresses account for cloud servers
that will be used for storing data and sends blocks of data to those hosts. An edge
server maintains a copy of each device’s identity and helps each device generate
homomorphic verifiable tags and data shards. In the prototype system, edge com­
puting processors are located close to the IoT devices in order to preprocess large
amounts of IoT data. This has the potential to significantly reduce the communication
costs and computational load associated with the system. Traditionally, P2P solutions
were based on client-server architecture where the server had powerful processing
abilities.

3.11.3.3 Big Data Storage

IoT devices can generate massive amounts of data, and Big Data technologies are
well-suited for handling and processing such large-scale data. A distributed and scal­
able data storage system is typically used to store IoT data in Big Data environments.
Distributed file systems like Hadoop Distributed File System (HDFS) are designed
to save huge amounts of data on a number of servers. HDFS provides fault tolerance,
data replication, and high throughput, making it suitable for IoT data storage. NoSQL
Databases: Various NoSQL databases like Apache Cassandra, MongoDB, or Apache
Unlocking the Power of IoT: An In-Depth Exploration 55

HBase are commonly used for IoT data storage. In addition to handling unstructured
and semi-structured data, NoSQL databases also handle IoT data [43].

3.11.4 Data Analytics and Insights in IOT

A data analytics approach involves analyzing both quantitative and qualitative data
about a digital property, in order to draw meaningful conclusions from them. Sensor/
Textual data produce a huge number of real-time datasets. As a result of the lack of
decentralized control, the management of IoT-based sensors is a complex process,
and aggregating data can be difficult. Due to their dynamic properties and limited
computational power, IoT sensor systems can benefit from using swarm intelligence
to solve complex problems. These datasets are preprocessed to utilize deep learning
architecture and swarm intelligence for analytics. In addition to providing a com­
plex problem solution using GPU-based memory, Tata Consultancy Service (TCS)
performs parallel analyzes of streaming data from multiple sensors 200 times faster
than traditional methods [44]. It processes huge amounts of sensor data and generates
useful insights. Social media data analytics utilize natural language processing and
text mining in order to extract useful knowledge. Convolutional neural networks and
video content analysis were used to represent useful data from image/video data.
Through spatial, temporal visualization and location intelligence, geospatial data
analytics can help us explore high-resolution images and data. A geospatial intel­
ligence application can provide unique insights, reveal hidden patterns, and deliver
information for better decision-making based on geospatial data [45]. Analyzing all
of these data sets in an automated manner will be possible with data analytics.

3.11.5 Security and Privacy in IOT

Digital technologies, namely artificial intelligence, blockchain, and machine
learning have been used to improve the security of data collected at the device
level. IoT device data currently is not protected by current standards, resulting in
individual confidentiality issues [46]. There are many inconsistencies and gaps
in the current standards for securing IoT data, and they contribute to inconsist­
encies and gaps in future standards. Examples of typical dataflow management
applications in smart manufacturing, smart transportation, and smart cities are
presented to illustrate the practical benefits of efficient IoT dataflow manage­
ment [47]. Transporting encrypted data at rest or in motion using standards such
as Secure Socket Layer (SSL), Federal Information Protection Standard (FIPS),
Secure Socket Shell (SSH), and ISO 27001 is used in communication middleware.
IoT devices have a few basic level security measures based on role-based access
control, encryption with regular updates and patches, and granular permissions.
The development of smart and intelligent sensors and actuators, as well as RFID
tags, has forward to the creation of a huge number of wireless networks with smart
and intelligent equipment (things) connected to the Internet constantly transmit­
ting data. The IoT makes it extremely challenging to maintain privacy and security
for all this data. Therefore, numerous present and future applications must priori­
tize safety and security.
56 IoT and Machine Learning for Smart Applications

3.11.5.1 Introduction
Networking and connectivity are fundamental pillars of the IoT ecosystem. The
IoT devices are constructed on a system with interconnected gadgets that can net­
work and share data instantly. IoT networking and connectivity encompass a wide
range of technologies, protocols, and architectures that enable efficient and secure
communication among IoT devices and cloud-based services. Networking refers
to the integration of computing devices to share resources, data, and information.
It allows computers, servers, printers, and other devices to communicate locally
and internationally (through the Internet). Networking is an essential component
of modern technology, facilitating communication, and data transfer between
devices.
Connectivity in IoT is a vast field with various subdisciplines, including net­
work administration, network security, network design, and network engineering.
For various types of networking, connectivity is also used in the IOT applica­
tion. The IoT relies heavily on networking to enable communication and data
exchange between a vast number of connected devices. IoT refers to a concept in
which everyday objects or devices are arranged through sensors, actuators, and
communications capabilities, allowing them to collect and exchange data via the
Internet or other networks. From smart home appliances to wearable technology to
industrial machinery and smart city infrastructure, these interconnected devices can
serve a variety of purposes.
In today’s interconnected world, it is critical for enterprises, organizations, and
individuals to understand networking fundamentals in order to facilitate efficient
and secure communication. Typically, network protocols describe how data is trans­
mitted over a computer network according to a set of rules and conventions. Data
packet protocols define not only the format and structure of data packets, but also
the procedures involved in creating, maintaining, and terminating communication
sessions between devices connected to a network, as well as the structure of the
data packets themselves. These are the physical devices used in a network, such as
computers, routers, switches, access points, and servers.
There are several key components and concepts in networking devices:

• Protocols: Transmitting and receiving data is controlled by protocols, which

are rules that govern the transmission and reception of data. The purpose of a
network protocol is to facilitate communication between computers that are
connected to a local area network (LAN), a wide area network (WAN), or the
Internet. They ensure that devices from different manufacturers and running
different software can communicate effectively and reliably.
a. LAN: In computer networking, a LAN is a network of devices that
are interconnected within a small area, usually within a building or
a household. The Internet, printers, and files are commonly shared
over LANs.
b. WAN: The WAN connects LANs and other networks across long
distances and geographically dispersed areas. WANs are most commonly
associated with the Internet.
Unlocking the Power of IoT: An In-Depth Exploration 57

• Router: A router is an electronic device that connects several networks and

sends data packets between them. To transport data from a source to a destin­
ation, it determines the best route.
• Switch: LAN switches are devices used to connect multiple devices. It
uses MAC addresses to determine where to forward data within the local
network.
• Firewall: According to predefined security rules, a firewall analyzes and
prevents network traffic. As a result, unauthorized access and other threats are
prevented from entering the network.
• Wireless Networking: Using radio waves, wireless networks allow devices to
be connected without having to rely on physical cables to do so. Wi-Fi exem­
plifies wireless networking.
• Internet Protocol (IP) Address: Protocols such as this one are responsible for
routing and addressing data packets in order for them to be properly delivered
across different networks. An Internet Protocol address is a number designa­
tion that is unique to each device on a network. It enables devices to connect
and interact with one another via an IP-based network.
• Domain Name System (DNS): A domain name is converted into an IP
address that computers can use to find each other on the Internet by converting
human-readable characters into numbers.
• A DNS is a protocol for translating human domain names into IP addresses.
Rather than entering a numerical IP address, users can access websites using
easy-to-remember names.

Networking in IoT is a multidimensional challenge that requires considerations for

scalability, security, energy efficiency, and interoperability to create robust and reli­
able ecosystems of connected devices.

3.11.6 Wireless Technologies for IOT

Wireless technologies enable seamless communication between devices without phys­
ical connections, which is essential for the widespread adoption of IoT applications.
As technology continues to advance, IoT networking will continue to evolve to meet
the demands of various industries and use cases. IoT devices utilize various commu­
nication protocols to establish connections and exchange data [48]. There are several
wireless technologies specifically designed to meet the requirements of IoT devices,
each with its strengths and use cases. Here are some of the prominent wireless tech­
nologies used in IoT.

Wireless Connectivity: Most IoT devices utilize wireless communication technolo­

gies due to their flexibility and ease of deployment. Some common wireless connect­
ivity options in IoT include:

Wi-Fi: Wireless technology such as Wi-Fi allows fast data transfer between devices
and the Internet, making it ideal for applications requiring fast communication. Data
58 IoT and Machine Learning for Smart Applications

can be transferred over LANs at high speeds, making it suitable for home, office, and
industrial settings where Wi-Fi access points are nearby.

Bluetooth: Designed for short-range communications between devices with low

power consumption, Bluetooth is the ideal technology for this application. A wide
variety of wearable devices, smart home gadgets, and health monitoring applications
make use of this technology.

Zigbee: Wireless communication protocols such as Zigbee are designed for devices
to communicate at low power, low rate, and short distances. This technology is pri­
marily used to create wireless sensor and control networks, also called the “Internet
of Things” (IoT). As part of IEEE 802.15.4, Zigbee provides a framework for creating
reliable and scalable wireless networks. The system is generally used in smart lighting
applications, industrial monitoring applications, and home automation applications.

Z-Wave: In addition, Z-Wave is another wireless protocol that is capable of supporting

home automation and smart homes. Unlike other smart home networks, it operates in
the sub-GHz frequency range, which provides good range and reliability. This type
of network protocol is commonly used for smart home applications, particularly for
home automation and security systems.

Cellular Connectivity: IoT devices can leverage existing cellular networks (2G, 3G,
4G, and now transitioning to 5G) to connect to the Internet over long distances. Cellular
connectivity is particularly suitable for IoT deployments in remote areas or applications
requiring wide-area coverage, such as asset tracking and smart city solutions.

Low-Power WAN (LPWAN): This technique caters specifically to IoT devices

that require long-range connectivity while operating on low power. It operates
on a licensed spectrum, offering better security and reliability compared to some
unlicensed LPWAN technologies.
Examples include:

• LoRaWAN: The technology is ideal for applications such as smart cities,

agriculture, and environmental monitoring due to its long range and low
power consumption.
• NB-IoT (Narrowband IoT): IoT deployments requiring low-power and
wide-area communication using cellular-based LPWANs.
• Sigfox: Sigfox is an LPWAN technology that operates in the unlicensed spec­
trum and is known for its simplicity and low-cost connectivity. It is suitable
for applications such as asset tracking, environmental monitoring, and smart
city management. A proprietary LPWAN technology known for its simplicity
and low-energy requirements.

Wired Connectivity: Though less common in many consumer IoT applications, some
devices may utilize wired connectivity, such as Ethernet, to connect to the Internet or
local networks. This approach is more prevalent in IIoT settings.
Unlocking the Power of IoT: An In-Depth Exploration 59

Mesh Networking: Mesh networks are formed when IoT devices communicate
with each other directly, rather than relying solely on a centralized hub or access
point. Mesh networking is advantageous for its self-healing capabilities and extended
coverage range. Zigbee and Thread are examples of protocols used to create mesh
networks.

Edge Computing: IoT generates a large amount of data, which is difficult to transfer
efficiently to centralized servers due to the large amount of data generated. In order to
conserve network bandwidth and reduce latency, edge computing is used to process
and analyze data at the edge (i.e., the IoT device) in order to process and analyze data.

Security: Security concerns arise from IoT connectivity. Implementing secure

authentication, encryption, and data integrity measures on devices can prevent cyber
threats and unauthorized access to sensitive data.

Interoperability: A cohesive IoT ecosystem requires IoT devices to be able

to communicate despite differences in manufacturers, platforms, or protocols.
Standardization efforts and common communication protocols contribute to
achieving interoperability.
Communication range, power requirements, data rate, deployment environment,
and cost are factors that must be considered when selecting a wireless technology
for an IoT application. As IoT continues to evolve, new wireless technologies and
improvements in existing ones are likely to emerge, further expanding the possibil­
ities for IoT deployments.

3.11.7 Cellular Networks and IOT

IoT devices need cellular networks to communicate and connect across large geo­
graphical areas. Suitable for long-range communications and seamless connectivity,
these devices deliver reliable, secure, and high-speed data transmission. It involves
the use of the same technology that powers your smartphone to connect physical
objects (such as sensors) to the Internet. IoT devices do not require a separate, private
network, but can instead be connected to the same mobile network as smartphones.
There is a possibility that cellular IoT is able to provide an alternative to non-cellular
LPWAN, like LoRaWAN or Sigfox.
Cellular networks running on 4G provide a broadband connection for voice calls
and video streaming applications in the consumer mobile markets. Yet, they are
pretty costly. They demand a lot of electricity, and it can be difficult to provide partial
inside coverage in industrial settings. They do, however, make a fantastic choice for
LPWAN networks. Ultimately, IoT data must be sent through the Internet, with 4G
being the best global choice for capacity, scalability, and compatibility [49]. Recent
news coverage of 5G has been mixed, with positive and negative aspects. The tech­
nology, supports mobility at extreme speeds with extremely low latency. Self-driving
cars and augmented reality are expected to use this technology in the future.
Wide Area Coverage is one of the primary advantages of using cellular networks
for IoT is their extensive coverage. Cellular infrastructure is already well-established
60 IoT and Machine Learning for Smart Applications

in many regions, providing connectivity even in remote areas. This wide area
coverage allows IoT devices to operate in various locations, making them suitable for
applications such as environmental monitoring, smart agriculture, and asset tracking.

• High Data Transfer Rates: In the last few years, cellular technologies have
dramatically improved the transfer of data from 3G to 4G and now to 5G.
This improvement in bandwidth enables real-time communication and high-
throughput applications, making cellular networks suitable for video surveil­
lance, industrial automation, and healthcare monitoring.
• Reliability and QoS: Cellular networks are designed to offer high reliability
and consistent connectivity. Unlike smart grids, transportation systems, and
emergency services, which are dependent on IoT, this is essential for mission-
critical IoT applications, where downtime or communication interruptions
can have severe consequences.
• Security: A cellular network protects the data that is transmitted between IoT
devices and the cloud, including encryption and authentication, as well as
other security components. This level of security is critical for safeguarding
sensitive information and preventing unauthorized access to IoT systems.
• Global Roaming: For IoT applications with international deployments, cel­
lular networks support global roaming capabilities. This allows devices to
connect seamlessly to different networks in various countries, enabling inter­
national tracking, logistics, and asset management solutions.
• Cost-Efficiency: With the widespread adoption of cellular technologies,
the cost of connectivity has become more affordable. Many cellular net­
work providers offer cost-effective data plans and packages tailored for IoT
deployments, making it more accessible for businesses and organizations to
implement large-scale IoT solutions.
• Mobility Support: Cellular networks support mobility, allowing IoT devices
like vehicles and wearable devices to maintain continuous connectivity while
on the move. This capability is essential for applications like connected cars,
fleet management, and healthcare wearables.
• Future-Proofing with 5G: The emergence of 5G technology brings signifi­
cant enhancements to cellular networks, such as lower latency, higher data
rates, increased device density, and support for massive machine-type com­
munication (mMTC). 5G’s attributes will unlock new possibilities for IoT
applications, particularly in areas requiring real-time responsiveness, like
augmented reality, industrial automation, and smart cities.

It was recommended for submission for consideration as an International Mobile

Telecommunications (IMT)-2020 standard by the wireless industry standardiza­
tion body 3rd Generation Partnership Project (3GPP), and it is capable of delivering
significant increases in gigabit rates for both uploads and downloads. Despite the
advantages, cellular networks do have some limitations for certain IoT use cases. The
reliance on cellular infrastructure can result in higher power consumption for IoT
devices, which may be a concern for battery-operated devices that require long-term
Unlocking the Power of IoT: An In-Depth Exploration 61

deployments [50]. Also, LPWAN or satellite communications might be a better option

in areas with low cellular coverage. Overall, cellular networks remain a significant
enabler for IoT, offering reliable, scalable, and secure connectivity for a wide range of
submissions across businesses. Mobile networks will continue to play an increasingly
important role in supporting IoT ecosystems as technology evolves.

3.11.8 Short-Range Communication in IOT

IoT devices can exchange data over relatively close distances with each other and
with other IoT devices using short-range communications. These technologies are
particularly useful in scenarios where devices need to interact within a confined space
or when they are in proximity to each other. Short-range communication technolo­
gies are especially useful in scenarios where long-range connectivity is unneces­
sary, impractical, or consumes too much power. These devices are primarily used in
buildings, homes, factories, and other environments in which devices must interact
closely with one another.
One of the significant advantages of short-range communication is its lower power
consumption, which allows for prolonged battery life in IoT devices. Short-range
technologies often have lower implementation costs, making them more accessible
for various IoT deployments. In order to achieve the greatest level of IoT integration
with other devices and platforms, it’s important to select the right technology for
short-range data transfer [51]. As the IoT ecosystem continues to evolve, short-range
communication technologies will remain vital components, complementing long-
range connectivity options and driving innovation across various industries.
When using RFID, which transmits small quantities of data, a reader and an RFID
tag can be separated by just a few centimeters. Technology has been used in the retail
and logistics sectors. A variety of items and machinery can be embedded with RFID
tags to track inventory and assets in real-time. Improved supply chain management
and better stock planning are possible as a result. Self-checkout, smart mirrors, and
other IoT applications are made possible by RFID’s continued entrenchment in the
retail industry, which is growing IoT adoption.

3.11.9 IOT Gateways and Edge Devices

As an intermediary between IoT devices deployed in the field and the cloud, an
IoT gateway is responsible for bridging the gap between these two environments.
It connects various sensors and devices to the cloud for processing and analysis
after the data has been aggregated. An IoT architecture cannot be completed without
gateways and edge devices. These platforms facilitate efficient communication, data
processing, and connectivity between IoT devices and cloud-based platforms. In the
context of the IoT, it is important to keep in mind that many devices use different com­
munication protocols and technologies. As a result of translating and bridging these
protocols, the gateway is able to ensure seamless communication and data exchange
across heterogeneous devices and the cloud. Figure 3.4 mentions the different IoT
gateways, which come in a variety of shapes, sizes, and types.
62 IoT and Machine Learning for Smart Applications

FIGURE 3.4 IoT gateways come in a variety of shapes, sizes, and types.

The success of IoT implementations depends on both IoT gateways and edge devices.
The use of IoT gateways facilitates communication between devices and the cloud by
facilitating data flow between them [52]. Edge devices, on the other hand, serve as the
frontline data collectors and processors, bringing real-time capabilities and localized intel­
ligence to the IoT ecosystem. Together, they form a robust and responsive IoT infrastruc­
ture that addresses the unique challenges and requirements of various IoT applications.
In addition to interconnecting sensing networks, gateways are used to connect cloud
computing environments or data centers to the Internet. Its objective is to manage het­
erogeneous data collected by multiple devices in many forms and deliver this data to
a developed phase [53]. Before transferring assembled data to data centers for optimal
operation and management of IoT devices, it should be cleaned, pre-processed, and
filtered. In terms of functionality, gateways can be divided into two types: basic
gateways and smart gates. By forwarding data incoming from low-end IoT devices to
data centers, the basic gateway serves as a gateway between them. In contrast, a smart
gateway performs data processing efficiently by preprocessing, filtering, and analyzing
the data prior to distributing it to the cloud database. The purpose of these intermediary
gadgets is to function in harsh environments and improve from failures while restoring
the displacement gap in the shortest amount of time possible.
As a gateway, software, and hardware equipment are used together. It consists of
a small software package that manages, preprocesses, and stores nodes in a network.
Additionally, a modest backup should be included to handle the connection defeat
issue, which involves maintaining the current configuration, running it in nested
mode for restoration, and performing data restoration from the same location.

3.11.10 IOT Network Protocols and Standards

Devices, gateways, and cloud platforms must work together in order to communicate
efficiently. An IoT network protocol and standard facilitate this exchange of data and
interoperability. These protocols and standards ensure that IoT systems can work
seamlessly together, regardless of the manufacturers or technologies involved.
Some of the real IOT application information needs to be exchanged in the
various protocols, making it popular for M2M communication. A lightweight MQTT,
Unlocking the Power of IoT: An In-Depth Exploration 63

publish-subscribe messaging protocol caters to low-bandwidth, high-latency, and

unreliable networks [54]. In order to operate on devices and networks with limited
resources, CoAP is developed to be a lightweight application layer protocol. As part
of the IoT, it is commonly used in applications where devices have limited processing
power and memory, namely IoT smart homes and IoT industrial automation systems.
Among IoT devices and cloud-based platforms, HTTP is widely used for communi­
cation. As IoT devices are able to communicate with RESTful APIs via HTTP, they
can be easily added to existing web applications and web services. A Bluetooth Low
Energy (BLE) protocol is a wireless communication protocol that is widely used in
IoT, especially in wearable devices, healthcare applications, and smart home systems.
IoT mesh networks can be created using ZigBee devices, which provide low-power,
low-data-rate wireless communication, particularly for industrial and smart home
applications. Device management protocols such as LwM2M are designed to help
IoT devices and applications to be managed efficiently.
As new technologies and innovations emerge, the IoT network protocol and
standard landscape continue to evolve. It is often difficult to determine which protocol
is most appropriate depending on the specific IoT use case, device constraints, data
transfer requirements, and the current ecosystem of platforms and devices [55].
To facilitate seamless communication between IoT devices and services, unified
standards and protocols are being developed to facilitate interoperability between
different protocols and standards.

REFERENCES
1. Jianxin Wang, Ming K. Lim, Chao Wang and Ming-Lang Tseng, “The Evolution
of the Internet of Things (IoT) over the Past 20 Years,” Computers & Industrial
Engineering, vol. 155, p. 107174, 2021, ISSN 0360-8352, https://doi.org/10.1016/
j.cie.2021.107174.
2. Anass Sedrati, Abdellatif Mezrioui and Aafaf Ouaddah, “IoT-Gov: A Structured
Framework for Internet of Things Governance,” Computer Networks, vol. 233,
p. 109902, 2023, ISSN 1389-1286, https://doi.org/10.1016/j.comnet.2023.109902 .
3. Colakovic A., Hadzialic, M., “Internet of things (loT): A Review of Enabling
Technologies, Challenges, and Open Research Issues,” Computer Networks, vol. 144,
pp. 17-39, Oct. 2018, https://doi.org/10.1016/j.comnet.2018.07.017.
4. Yosra Hajjaji, Wadii Boulila, Imed Riadh Farah, Imed Romdhani and Amir Hussain,
“Big Data and IoT-based Applications in Smart Environments: A Systematic Review,”
Computer Science Review, vol. 39, p. 100318, 2021, ISSN 1574-0137, https://doi.org/
10.1016/j.cosrev.2020.100318.
5. M. Z. Khan, O. H. Alhazmi, M. A. Javed, H. Ghandorh and K. S. Aloufi, “Reliable
Internet of Things: Challenges and Future Trends,” Electronics, vol. 10, no.19,
pp. 23-77, Sep. 2021, https://doi.org/10.3390/electronics10192377.
6. Kehinde Lawal and Hamed Nabizadeh Rafsanjani, “Trends, Benefits, Risks, and
Challenges of IoT Implementation in Residential and Commercial Buildings,” Energy
and Built Environment, vol. 3, no. 3, pp. 251-266, 2022, ISSN 2666-1233, https://doi.
org/10.1016/j.enbenv.2021.01.009.
7. Ibrar Yaqoob, Khaled Salah, Raja Jayaraman and Mohammed Omar, “Metaverse
Applications in Smart Cities: Enabling Technologies, Opportunities, Challenges, and
64 IoT and Machine Learning for Smart Applications

Future Directions,” Internet of Things, vol. 11, p. 100884, 2023, ISSN 2542-6605,
https://doi.org/10.1016/j.iot.2023.100884.
8. Mirani, G. Velasco-Hernandez, A. Awasthi and J. Walsh, “Key Challenges and
Emerging Technologies in Industrial IoT Architectures: A Review,” Sensors, vol. 22,
no. 15, p. 5836, Aug. 2022, https://doi.org/10.3390/s22155836.
9. Nishant Chaurasia and Prashant Kumar, “A Comprehensive Study on Issues and
Challenges Related to Privacy and Security in loT,” e-Prime - Advances in Electrical
Engineering, Electronics and Energy, vol. 4, p. 100158, 2023, ISSN 2772-6711,
https://doi.org/10.1016/j.prime.2023.100158.
10. Evolution of Internet of Things. www.techaheadcorp.com/knowledge-center/evolut
ion-of-iot/, accessed 03 August 2023.
11. Security, privacy and trust of different layers in Internet-of-Things (IoTs) framework
by A. Tewari and B.B. Gupta.
12. M. M. Islam, S. Nooruddin, and F. Karray, “Internet of Things: Device Capabilities,
Architectures, Protocols, and Smart Applications in Healthcare Domain,” IEEE
Internet of Things Journal, vol. 7, no. 10, pp. 10093-10108, Oct. 2020.
13. H. A. Khattak, M. A. Shah, S. Khan, I. Ali, and M. Imran, “Perception Layer Security
in Internet of Things,” Future Generation Computer Systems, vol. 100, pp. 144-164,
Nov. 2019.
14. M. Lombardi, F. Pascale, and D. Santaniello, “Internet of Things: A General Overview
between Architectures, Protocols and Applications,” Information Systems Frontiers,
vol. 23, pp. 1261-1281, Mar. 2021.
15. B. Padma and S. B. Erukala, “Peer-to-Peer Communication Protocol in IoT-Enabled
ZigBee Network: Investigation and Performance Analysis,” Journal of Ambient
Intelligence and Humanized Computing, vol. 12, pp. 7497-7505, 2021.
16. C. K. Rath, A. K. Mandal, and A. Sarkar, “Microservice Based Scalable IoT Architecture
for Device Interoperability,” in Proceedings ofthe 2020 IEEE International Conference
on Advanced Networks and Telecommunications Systems (ANTS), 2020, pp. 1-6.
17. A. Gerodimos, L. Maglaras, M. A. Ferrag, N. Ayres, and I. Kantzavelou,
“IoT: Communication Protocols and Security Threats,” in Proceedings of the 2019
IEEE Global Communications Conference (GLOBECOM), 2019, pp. 1-6.
18. R. Attarian and S. Hashem, “An Anonymity Communication Protocol for Security and
Privacy of Clients in IoT-Based Mobile Health Transactions,” IEEE Transactions on
Industrial Informatics, vol. 16, no. 4, pp. 2702-2711, Apr. 2020.
19. C. Bayilmis, M. A. Ebleme, U. Qavu^oglu, K. Ku^uk, and A. Sevina, “A Survey
on Communication Protocols and Performance Evaluations for Internet of Things,”
Internet of Things, vol. 9, p. 100149, Sep. 2020.
20. T. Dommguez-Bolano, O. Campos, V. Barral, C. J. Escudero, and J. A. Garrfa-
Naya, “An Overview of IoT Architectures, Technologies, and Existing Open-Source
Projects,” Electronics, vol. 9, no. 2, p. 275, Feb. 2020.
21. J. Trevathan, W. Read, and A. Sattar, “Implementation and Calibration of an IoT Light
Attenuation Turbidity Sensor,” Sensors, vol. 19, no. 9, p. 2130, May 2019.
22. J. D. A. Correa, A. S. R. Pinto, and C. Montez, “Lossy Data Compression for IoT
Sensors: A Review,” IEEE Access, vol. 8, pp. 68506-68527, 2020.
23. M. Kumar, “A Survey on Event Detection Approaches for Sensor Based IoT,” Journal
of King Saud University- Computer and Information Sciences, vol. 33, no. 5, pp. 489­
500, Jun. 2021.
24. F. Fama, J. N. Faria, and D. Portugal, “An loT-Based Interoperable Architecture for
Wireless Biomonitoring of Patients with Sensor Patches,” Sensors, vol. 20, no. 18,
p. 5266, Sep. 2020.
Unlocking the Power of IoT: An In-Depth Exploration 65

25. I. Nassra and J. V. Capella, “Data Compression Techniques in IoT-Enabled Wireless

Body Sensor Networks: A Systematic Literature Review and Research Trends for QoS
Improvement,” IEEE Access, vol. 8, pp. 61980-62000, 2020.
26. W. N. Hussein, “A Proposed Framework for Healthcare Based on Cloud Computing
and IoT Applications,” Journal of King Saud University- Computer and Information
Sciences, vol. 33, no. 7, pp. 918-924, Aug. 2021.
27. A. Rahman, M. J. Islam, S. S. G. Muhammad, K. Hasan, and P. Tiwari, “Towards a
Blockchain-SDN-Based Secure Architecture for Cloud Computing in Smart Industrial
IoT,” IEEE Internet of Things Journal, vol. 8, no. 4, pp. 2336-2344, Feb. 2021.
28. C. Yang and H. Ming, “Detection of Sports Energy Consumption Based on IoTs and
Cloud Computing,” Computers, Materials & Continua, vol. 68, no. 3, pp. 3177­
3190, 2021.
29. R. Bose, “Design of Smart Inventory Management System for Construction Sector
Based on IoT and Cloud Computing,” Wireless Personal Communications, vol. 118,
no. 1, pp. 541-557, May 2021.
30. G. Peng, “Constrained Multi Objective Optimization for IoT-Enabled Computation
Offloading in Collaborative Edge and Cloud Computing,” Computers & Electrical
Engineering, vol. 93, p. 107260, Oct. 2021.
31. M. Wang and Q. Zhang, “Optimized Data Storage Algorithm of IoT Based on Cloud
Computing in Distributed System,” IEEE Access, vol. 8, pp. 196695-196705, 2020.
32. S. Shen, “Optimal Privacy Preservation Strategies with Signalling Q-learning for
Edge-Computing-Based IoT Resource Grant Systems,” Computers & Electrical
Engineering, vol. 92, p. 107168, Sep. 2021.
33. J. Li, “Service Home Identification of Multiple-Source IoT Applications in Edge
Computing,” Sensors, vol. 21, no. 12, p. 4125, Jun. 2021.
34. M. Goudarzi, “An Application Placement Technique for Concurrent IoT Applications
in Edge and Fog Computing Environments,” Journal of Network and Computer
Applications, vol. 182, p. 103016, Dec. 2021.
35. Y. Chen, “Channel-Reserved Medium Access Control for Edge Computing Based
IoT,” IEEE Access, vol. 8, pp. 219297-219305, 2020.
36. W. Yu, “A Survey on the Edge Computing for the Internet of Thing,” IEEE Access,
vol. 8, pp. 102161-102174, 2020.
37. S. Chen, X. Zhu, H. Zhang, C. Zhao, G. Yang and K. Wang, “Efficient Privacy
Preserving Data Collection and Computation Offloading for Fog-Assisted IoT,” IEEE
Transactions on Sustainable Computing, vol. 5, no. 4, pp. 526-540, 1 Oct.-Dec. 2020,
https://doi.org/10.1109/TSUSC.2020.2968589 .
38. P. Ta-Shma, A. Akbar, G. Gerson-Golan, G. Hadash, F. Carrez and K. Moessner, “An
Ingestion and Analytics Architecture for IoT Applied to Smart City Use Cases,” IEEE
Internet of Things Journal, vol. 5, no. 2, pp. 765-774, Apr. 2018, https://doi.org/
10.1109/JIOT.2017.2722378.
39. J.-S. Fu, Y. Liu, H.-C. Chao, B. K. Bhargava and Z.-J. Zhang, “Secure Data Storage and
Searching for Industrial IoT by Integrating Fog Computing and Cloud Computing,”
IEEE Transactions on Industrial Informatics, vol. 14, no. 10, pp. 4519-4528, Oct.
2018, https://doi.org/10.1109/TII.2018.2793350 .
40. Q.-T. Doan, A. S. M. Kayes, W. Rahayu and K. Nguyen, “Integration of IoT Streaming
Data With Efficient Indexing and Storage Optimization,” IEEE Access, vol. 8,
pp. 47456-47467, 2020, https://doi.org/10.1109/ACCESS.2020.2980006.
41. Y. Tao, P. Xu and H. Jin, “Secure Data Sharing and Search for Cloud-Edge-
Collaborative Storage,” IEEE Access, vol. 8, pp. 15963-15972, 2020, https://doi.org/
10.1109/ACCESS.2019.2962600.
66 IoT and Machine Learning for Smart Applications

42. H. Wang and J. Zhang, “Blockchain Based Data Integrity Verification for Large-Scale
loT Data,” IEEE Access, vol. 7, pp. 164996-165006, 2019, https://doi.org/10.1109/
ACCESS.2019.2952635.
43. M. Babar, M. A. Jan, X. He, M. U. Tariq, S. Mastorakis and R. Alturki, “An Optimized
IoT-Enabled Big Data Analytics Architecture for Edge-Cloud Computing,” IEEE
Internet of Things Journal, vol. 10, no. 5, pp. 3995-4005, 1 Mar. 2023, https://doi.org/
10.1109/JIOT.2022.3157552.
44. A. R. Dargazany, P. Stegagno, and K. Mankodiya, “WearableDL: Wearable Internet-
of-Things and Deep Learning for Big Data Analytics—Concept, Literature, and
Future,” Mobile Information Systems, no. 1, 8125126.
45. S. A. Shah, D. Z. Seker, S. Hameed and D. Draheim, “The Rising Role of Big
Data Analytics and IoT in Disaster Management: Recent Advances, Taxonomy and
Prospects,” IEEE Access, vol. 7, pp. 54595-54614, 2019, https://doi.org/10.1109/ACC
ESS.2019.2913340.
46. D. Wei et al., “Dataflow Management in the Internet of Things: Sensing, Control, and
Security,” Tsinghua Science and Technology, vol. 26, no. 6, pp. 918-930, Dec. 2021,
https://doi.org/10.26599/TST.2021.9010029.
47. J. Cook, S. U. Rehman and M. A. Khan, “Security and Privacy for Low Power IoT
Devices on 5G and Beyond Networks: Challenges and Future Directions,” IEEEAccess,
vol. 11, pp. 39295-39317, 2023, https://doi.org/10.1109/ACCESS.2023.3268064.
48. Gunjan Beniwal and Anita Singhrova, “A Systematic Literature Review on IoT
Gateways”, Journal of King Saud University - Computer and Information Sciences,
vol. 34, no. 10, Part B, 2022, pp. 9541-9563, ISSN 1319-1578, https://doi.org/
10.1016/j.jksuci.2021.11.007.
49. Vaigandla, Karthik, Radha, Krishna and Allanki, Sanyasi Rao, A Study on IoT
Technologies, Standards and Protocols. 2021. https://doi.org/10.17697/ibmrd/2021/
v10i2/166798.
50. Pons, Mario, Estuardo Valenzuela, Brandon Rodriguez, Juan Arturo Nolazco-
Flores and Carolina Del-Valle-Soto, “Utilization of 5G Technologies in IoT
Applications: Current Limitations by Interference and Network Optimization
Difficulties—A Review”, Sensors, vol. 23, no. 8, p. 3876, 2023, https://doi.org/
10.3390/s23083876.
51. Tariq, Usman, Irfan Ahmed, Ali Kashif Bashir, and Kamran Shaukat, “A Critical
Cybersecurity Analysis and Future Research Directions for the Internet of Things: A
Comprehensive Review”, Sensors, vol. 23, no. 8, p. 4117, 2023, https://doi.org/10.3390/
s23084117.
52. Yaser Mohammed Al-Worafi, “Chapter 6 Internet Technologies for Drug Safety”,
Springer Science and Business Media LLC, 2023. https://doi.org/10.1007/978-3-031-
34268-4_6.
53. Isa Avci and Cevat Ozarpa. “Chapter 10 Machine Learning Applications and Security
Analysis in Smart Cities”, Springer Science and Business Media LLC, 2022. https://
doi.org/10.1007/978-3-030-97516-6_10.
54. Laguidi, Ahmed, Tamtam, Samiya and Mejdoub, Youssef. “A Technique to Improve
IoT Connectivity Based on NB-IoT and D2D Communications”, ITM Web of
Conferences, vol. 52, pp. 01010-01018, 2023, https://doi.org/10.1051/itmconf/2023
5201010.
55. Soori, Mohsen, Arezoo, Behrooz and Dastres, Roza. “Internet of Things for Smart
Factories in Industry 4.0, A Review”, Internet of Things and Cyber-Physical
Systems, vol. 3, pp. 192-204, 2023, https://doi.org/10.1016/j.iotcps.2023.04.006.
 Machine Learning
4 in Internet of Things
Sujni Paul and Grasha Jacob

4.1 INTRODUCTION
A new era of technological innovation has begun with the confluence of the Internet
of Things (IoT) and Machine Learning (ML), which has completely changed the
ways to gather, process, and use data from linked devices. A thorough investiga­
tion of this synergistic link is undertaken in this book chapter, revealing its signifi­
cant influence in a variety of fields. Fundamentally, the chapter seeks to clarify the
complex interactions between IoT and ML, providing readers with information and
understanding to help them navigate and take advantage of this revolutionary com­
bination. To ensure a strong foundation upon which to develop, a deep knowledge of
the fundamental ideas behind both ML and IoT is established.
The potential and problems that come along with the data of IoT, such as data var­
iety, velocity, volume, and accuracy are evaluated next. Additionally, ML techniques—
supervised, unsupervised, and reinforcement learning designed specifically for IoT
are explored to find how they may be used in various IoT scenarios. A range of fascin­
ating case studies and real-world applications will highlight the practical implications
of this integration, showing how ML, among other revolutionary use cases, optimizes
energy consumption, improves predictive maintenance procedures, and spurs innov­
ation in the healthcare industry. To enable readers to effectively utilize both tech­
nologies in their domains, this chapter aims to give a complete understanding of the
dynamic interplay between IoT and ML. It does this by providing information and
practical insights. The structure of the chapter is organized as follows.

4.2 FUNDAMENTALS OF IOT AND ML

An outline of the basic concepts of IoT and ML to establish a foundation for
understanding their convergence is provided in this section. It explains the archi­
tecture and components of IoT systems, including sensors, actuators, gateways, and
cloud platforms. The various techniques of ML with their capabilities and applications
are also introduced.

DOI: 10.1201/9781032623276-4 67
68 IoT and Machine Learning for Smart Applications

4.2.1 Overview of IOT Architecture and Components

IoT architecture is a complete framework built to orchestrate the integration and
functionality of varied components in the IoT ecosystem. At its foundation lies
the Perception Layer with sensors and actuators that capture real-world data. The
Network Layer ensures seamless connectivity through protocols like Wi-Fi and cel­
lular networks, while the Middleware Layer manages data processing, transformation,
and device communication. The processed data is then utilized in the Application
Layer, which houses specific use case applications that run from smart homes to
industrial automation. The Business Layer integrates IoT applications with broader
enterprise systems, aligning IoT data with overarching business objectives. Security
and Privacy Layer implements crucial measures such as authentication and encryp­
tion, safeguarding against unauthorized access. Cloud and Edge Computing Layer
involves storing and processing data either in the cloud or closer to the data source.
The User Interface Layer provides interfaces and dashboards for end-users to interact
with IoT systems. This intricate design enables the creation of scalable, efficient, and
secure IoT solutions tailored to diverse applications and industries [1].

4.2.2 Introduction to ML Techniques

ML is a transformative subdomain within the domain of Artificial Intelligence (AI)
that empowers computer systems to recognize patterns for making predictions or
decisions without being explicitly programmed. ML enables computers to acquire
knowledge from data and experiences, thereby improving their performance over
time [2]. This transition from traditional rule-based programming to learning from
data has opened up new potential across various domains.
One fundamental category of ML is supervised learning. In the case of supervised
learning, models have to be trained with labeled datasets, and the test data(input) is
then mapped onto the corresponding output labels by the algorithm. For instance, in
a supervised learning task of image recognition, the algorithm is trained on a dataset
of images with corresponding labels, learning to associate particular features on the
images with the correct labels.
Unsupervised learning deals with unlabeled data, looking for patterns or structures
within the dataset. Clustering is an unsupervised learning technique in which similar
data points are grouped together. An example of unsupervised learning is customer
segmentation, where the algorithm identifies groups of customers with analogous
purchasing activities without predefined categories.
Reinforcement learning is a technique that is focused on training agents. The agent
gets feedback in the form of penalties or rewards, thereby learning the best strat­
egies based on trial and error. This approach is particularly effective in scenarios such
as autonomous systems, robotics, and game playing. ML algorithms are classified
into several types—neural networks, decision trees, support vector machines (SVM),
and ensemble methods. Neural networks consist of nodes that are interconnected
to process information in layers. Decision trees are tree-like structures that make
decisions based on input features. SVMs are applied for classification and regression
tasks, finding the ideal hyperplane to separate different classes. Ensemble methods,
Machine Learning in Internet of Things 69

like random forests, combine multiple models to enhance overall performance. The
success of ML relies heavily on data quality, feature selection, and model evalu­
ation. Additionally, ethical considerations, interpretability, and fairness in ML models
have become more and more important. As ML advances, it revolutionizes automated
complex tasks, industries, and contributes to solving challenging problems in diverse
fields [3].

4.2.3 Convergence of IOT and ML

The convergence of the IoT and ML represents a powerful synergy that amplifies
the capabilities of both technologies. At its core, IoT involves the interconnection
of physical devices, sensors, and actuators embedded in everyday objects, enabling
them to gather and swap data. This vast network generates a massive volume of data
that, when properly harnessed, becomes an important resource for ML algorithms.
An important technical aspect of this convergence lies in the data-driven nature
of both IoT and ML. IoT devices continuously generate real-time data streams, pro­
viding a rich source of information about the environment, user behavior, and system
performance. ML algorithms can influence this data to uncover patterns, correlations,
and trends that may be challenging for traditional rule-based programming to dis­
cern. For instance, prognostic maintenance in industrial IoT can benefit from ML
algorithms that analyze sensor data to forecast equipment failures, optimize mainten­
ance schedules, and minimize downtime.
The incorporation of ML models within IoT systems introduces a layer of intelli­
gence that enables devices to adapt, learn, and make autonomous informed decisions.
Edge computing plays a vital role in this convergence by bringing ML capabilities
closer to the data source, reducing bandwidth requirements and latency. Deploying
lightweight ML models directly on IoT devices or edge gateways allows for imme­
diate response and real-time analysis, which is critical in applications like smart
cities, healthcare monitoring, and autonomous vehicles.
The merging of IoT and ML introduces challenges related to data privacy, energy
efficiency, and security. As sensible information is collected by IoT, ML models need
to be designed with privacy-preserving techniques, ensuring that personal data is
adequately protected. Security measures, such as encryption and secure communi­
cation protocols, become necessary to safeguard both the IoT infrastructure and the
ML algorithms.
Hence, the technical convergence of IoT and ML is transforming how we collect,
process, and derive insights from data. This integration empowers IoT systems to
evolve beyond simple data collection and act as intelligent entities capable of learning
from their surroundings. As both technologies continue to advance, their synergy
holds enormous prospects in creating novel solutions to diverse domains, from indus­
trial automation to smart homes and beyond [4].

4.3 CHALLENGES AND OPPORTUNITIES IN IOT DATA

The abundance of data that is generated by IoT devices presents both challenges
and opportunities. This section discusses the exceptional characteristics of IoT data,
70 IoT and Machine Learning for Smart Applications

including scalability, heterogeneity, and real-time processing requirements, along

with the challenges related to handling large volumes of data from diverse sources
and emphasizes the potential of ML to haul out valuable insights and patterns from
this data.

4.3.1 Data Heterogeneity and Scalability in IOT

One of the primary challenges in managing IoT data lies in its inherent heterogen­
eity. IoT devices generate diverse types of data, ranging from text and numerical
data to image and video feeds. Additionally, these devices often operate on different
communication protocols and standards, making it challenging to integrate and ana­
lyze data cohesively. This heterogeneity poses obstacles to ensuring interoperability
and seamless communication across the IoT ecosystem. Moreover, when the quantity
of devices connected increases, the volume of data spawned in excess necessitates
robust scalability solutions. Conventional data processing systems find it difficult to
work with the increasing influx of data streams, requiring the progress of scalable
and efficient infrastructure capable of accommodating the evolving demands of the
IoT landscape.

4.3.2 Real-Time Processing Requirements

IoT applications frequently demand real-time processing capabilities to mine appro­
priate insights and support instantaneous decision-making. In scenarios such as pre­
dictive maintenance or autonomous vehicles, delays in data processing can have
significant consequences. Meeting these real-time processing requirements involves
addressing latency challenges in data transmission, storage, and analytics. Inefficient
edge computing solutions allow data to be processed closer to the source instead
of relying exclusively on the centralized cloud servers, thereby helping to mitigate
latency concerns. This methodology reduces the time taken for data to travel and
alleviates network congestion, ensuring that critical decisions can be made promptly
based on the most current information available.

4.3.3 The Prospective of IOT Data for ML Applications

While IoT data presents challenges, it brings plenty of opportunities for ML
applications. The massive data produced by IoT devices serves as a valuable resource
for training and improving ML models. These models can be employed for anomaly
detection, predictive analytics, and pattern recognition, enhancing the overall func­
tionality of IoT systems. For instance, in smart homes, ML algorithms can observe
user behavior patterns to optimize energy consumption. However, leveraging this
potential requires addressing data quality issues, ensuring data privacy and security,
and developing ML models capable of handling the dynamic and evolving nature
of IoT data. The incorporation of ML with IoT enhances automation and decision­
making capabilities and unfold avenues for innovation across various industries.
In navigating the prospects and challenges of IoT data, it is vital for technologists and
businesses to adopt a holistic approach, considering factors like data standardization,
Machine Learning in Internet of Things 71

real-time processing capabilities, scalability, and the strategic integration of ML

applications. These aspects, when tackled, expose the full prospects of IoT, paving
the way to responsive and more intelligent systems.

4.4 ML TECHNIQUES FOR IOT

4.4.1 Supervised Learning in IOT

In IoT, supervised learning entails using labeled data—data for which the input­
output correlations are known—to train ML models. Using the given training dataset,
the algorithm in this paradigm learns to make predictions or judgments by mapping
input features to corresponding output labels. Thanks to this training process, the
model can generalize and make precise predictions on new, untested data. Supervised
learning is crucial for IoT applications like predictive maintenance, device behavior
classification, and anomaly detection. IoT applications can obtain insights into oper­
ational patterns, security concerns, and optimal resource utilization by utilizing
algorithms like decision trees, regression models, and SVM.
Supervised learning has an extensive range of uses in IoT, from boosting
cybersecurity to increasing industrial process efficiency. For instance, supervised
learning algorithms identify user behaviors in IoT-based smart homes to differen­
tiate between benign and possibly dangerous activities. In industrial environments,
sensor data is analyzed using predictive maintenance models created by supervised
learning to anticipate equipment breakdowns, cutting downtime, and mainten­
ance expenses. Supervised learning approaches enable IoT systems to make well-
informed decisions, hence enhancing their overall functionality and reliability in
multiple domains.

4.4.1.1 Classification and Regression Algorithms

• Decision trees (DT)

DT are an indispensable tool for categorizing IoT devices depending on their behavior.
DT can be utilized in the perspective of IoT security to distinguish between legit­
imate and malicious activity. For example, a decision tree model can detect anomalies
and possible security risks by examining the communication and behavior patterns
of devices. DTs have the benefit of being simply interpretable making it easier for
stakeholders to comprehend the decision-making process [5].

• SVM

SVM is an effective technique for categorizing IoT data, particularly in situations

involving IoT network intrusion detection. Through input data mapping into a high­
dimensional space, SVM determine the optimal hyperplane for classifying data. SVM
can be used in an IoT security setting to differentiate between typical and unusual net­
work data. Because of this feature, it’s a useful tool for securing IoT networks from
online attacks [6].
72 IoT and Machine Learning for Smart Applications

• Random Forest

Overfitting in IoT classification issues is a problem that can be effectively

addressed by using Random Forest, an ensemble learning technique. Higher accuracy
and resilience are achieved by the model through the combination of numerous deci­
sion trees. A Random Forest model can categorize sensor data produced by numerous
IoT devices in an application scenario, for example, predicting equipment failures
or identifying environmental conditions. Random Forest is an effective tool for pro­
cessing varied datasets that are typical in IoT scenarios because of its scalability and
versatility [7].

4.4.1.2 Use Cases and Applications

• Decision trees

By anticipating that equipment is likely to break, predictive maintenance, a crucial IoT

application, ensures peak performance and reduces downtime. Models that evaluate
sensor data and predict maintenance requirements can be developed using decision
trees. Now, let’s look at an example to show how a decision tree might be used in
an IoT context for predictive maintenance. Figure 4.1 shows the diagrammatical
representation of the decision tree.
The decision tree classifier is trained on simulated IoT sensor data, where ‘Sensor1’
and ‘Sensor2’ represent sensor readings, and ‘MaintenanceNeeded’ is a binary label
indicating whether maintenance is required. The decision tree is then visualized,
allowing interpretation of the decision-making process. This model can be extended
to any real-world IoT dataset for predictive maintenance in several industries like
manufacturing or healthcare.

• SVM—anomaly detection in network traffic

A useful application of SVM is finding anomalies in network traffic in IoT. SVMs

are used to differentiate normal and abnormal behavior in network communications
because they are good at finding patterns in data and categorizing them. The chart

FIGURE 4.1 Decision tree.

Machine Learning in Internet of Things 73

FIGURE 4.2 Confusion matrix.

2.20

2.15

2.10

2.05

2.00

1.95

- 1.90

- 1.85

i -1.80
Maintenance
Predicted

FIGURE 4.3 Predicted value.

given below illustrates anomaly identification by SVM for an IoT network. Figure 4.2
shows the confusion matrix for SVM.
In this example, the SVM classifier is trained on a simulated dataset where different
features represent network traffic characteristics, and ‘Anomaly’ is a binary label
indicating whether the traffic is normal or an anomaly. The SVM model is trained to
separate normal from anomalous traffic patterns. The confusion matrix visualization
provides insights into the model’s performance in detecting anomalies. Figure 4.3
shows the predicted value of SVM.
74 IoT and Machine Learning for Smart Applications

4.4.2 Unsupervised Learning in IOT

Through the usage of more intricate techniques like hierarchical, density-based
clustering, and more conventional methods like K-means clustering, readers will get
a deeper comprehension of how these algorithms cluster comparable data points and
expose underlying patterns and structures. The conversation also covers anomaly detec­
tion algorithms, demystifying methods like One-Class SVMs, and Isolation Forests.
For identifying anomalies in large and complex data sets produced by IoT devices and
to take preventative action against any problems, detection of anomaly is essential.
Here, the strategies for unsupervised learning, with a particular emphasis on
anomaly detection and grouping methods are emphasized. The investigation provides
readers with a systematic understanding of the revolutionary potential of unsuper­
vised learning in the IoT ecosystem by encompassing theoretical underpinnings, real-
world applications, and illustrative use cases.

4.4.2.1 Clustering and Anomaly Detection Algorithms

K-Means clustering algorithm is used to partition a dataset into K distinct, non­
overlapping subsets (clusters). This algorithm iteratively assigns each data point to
one of the K clusters depending on similar features. The objective is minimizing the
within-cluster sum of squares, meaning that the data points contained by each cluster
should be as close to each other as possible. The steps of the algorithm are as follows:

• Initialization: Randomly or strategically, K initial cluster centroids are

selected.
• Assignment: Each data point is assigned to the cluster whose centroid is
closest.
• Update: The centroids of the clusters are updated depending on the assigned
data points.
• Repeat: Assignment is iterated, and steps are updated until convergence.

K-Means is scalable and efficient, making it suitable for large datasets. The initial
choice of centroids is sensible and may converge to local minima.

4.4.3 Reinforcement Learning in IOT

From the perspective of IoT, reinforcement learning (RL) is a cutting-edge method for
enabling adaptive behavior and autonomous decision-making in networked devices.
In reinforcement learning (RL), an agent—typically an IoT device or system—learns
by interacting with its surroundings to make the best decisions. These choices fre­
quently have to do with resource management, energy efficiency, or adaptation to
dynamic situations in the framework of IoT. The agent receives feedback based on
its happenings in the form of incentives or penalties, which helps it to improve and
hone its decision-making techniques over time. RL has an extensive range of uses in
IoT, from resource allocation and traffic optimization in industrial settings to energy
management and predictive maintenance. When RL algorithms are incorporated into
IoT devices, there is potential for.
Machine Learning in Internet of Things 75

4.4.3.1 Learning from Interactions and Feedback

In ML, learning from encounters is a basic idea, especially when it comes to reinforce­
ment learning. According to this paradigm, an agent engages with its settings, acts,
and gets feedback in the form of incentives/ punishments. Learning a policy—a
collection of guidelines or tactics—that optimizes the cumulative reward over time
is the agent’s goal. Exploration and exploitation are key components of the learning
process, as the agent investigates various actions to determine their effects and uses
the knowledge at hand to make the best choices. By means of frequent encounters,
the agent enhances its comprehension of the surroundings, modifies its activities, and
ultimately elevates its capacity for making decisions.
For many applications, including recommendation algorithms and user inter­
face design, user involvement is essential. Utilizing data produced by user actions
to improve system and application performance is known as learning from user
interactions. In an online store, for instance, a recommendation engine can
deliver tailored recommendations based on consumers’ previous interactions with
products. With this strategy, systems may adjust to user behavior and personal
preferences, resulting in a more personalized and interesting experience. ML
algorithms provide insights that enhance user pleasure, boost productivity, and
facilitate better decision-making across a range of industries by examining trends
in user interactions.
Figure 4.4 chart visually represents how the cumulative rewards augment as the
agent is trained from its interactions over episodes.

FIGURE 4.4 Cumulative rewards vs episodes.

76 IoT and Machine Learning for Smart Applications

4.4.3.2 Use Cases and Applications

Energy management in smart buildings is an instance of a reinforcement learning use
case within IoT. Imagine an instance where a building is outfitted with an array of
IoT devices, including occupancy sensors, lighting controls, and smart thermostats.
Optimizing energy consumption and preserving occupant comfort is the goal. In this
situation, the RL agent might be a centralized control system that modifies HVAC
(Heating, Ventilation, and Air Conditioning) settings in response to outside environ­
mental factors and real-time occupancy trends. Figure 4.5 shows the Reinforcement
Learning and Energy Management.
Reinforcement Learning in IoT is applied in Industrial machinery predictive
maintenance. Consider a manufacturing plant whose equipment has a network
of IoT sensors built into it. Anticipating when equipment repair is required is
the aim in order to minimize downtime and lower maintenance expenses. In this
case, the RL agent can decide when to plan preventative maintenance by util­
izing historical data on machine health, sensor readings, and previous mainten­
ance activities.
In this example, the learning episodes are represented in the x-axis, and the
simulated maintenance predictions are represented in the y-axis. When the agent gets
trained from historical data interactions, sensor readings, and maintenance records,
the chart could demonstrate progress in the accuracy of maintenance predictions
over episodes. The goal is to effectively identify patterns indicative of impending
machine failures, ultimately optimizing the timing of preventive maintenance actions.
Figure 4.6 shows the Reinforcement Learning for Predictive Maintenance.

FIGURE 4.5 Reinforcement learning and energy management.

Machine Learning in Internet of Things 77

FIGURE 4.6 Reinforcement learning for predictive maintenance.

4.5 CASE STUDIES AND APPLICATIONS

To demonstrate the practical applications of ML in IoT, this section presents mul­
tiple case studies and real-world examples. It investigates how ML algorithms can
be leveraged for predictive maintenance in industrial IoT, assisting in the active
detection of potential breakdowns and optimizing conservation schedules. It further
explores the application of ML in anomaly detection in smart cities, facilitating the
early detection of unusual events or behaviors. Additionally, the section highlights the
role of ML in personalized healthcare monitoring, energy optimization in smart grids,
smart transportation, and agriculture and environmental monitoring.
Predictive maintenance in Industrial IoT involves the use of data analytics, ML,
and sensor technologies to predict when equipment or machinery is likely to fail,
and allowing for timely maintenance to prevent unplanned downtime and opti­
mize maintenance schedules. Let’s explore a real-time case study to illustrate this
concept.

CASE STUDY 1: PREDICTIVE MAINTENANCE IN

A MANUFACTURING PLANT
Scenario:
Imagine a large manufacturing plant that produces automotive components.
The plant relies on a variety of machines and equipment to carry out its
operations, and any unexpected breakdown can result in significant production
losses.
78 IoT and Machine Learning for Smart Applications

This implementation requires the following

Sensor Deployment:
The considered manufacturing plant is furnished with a network of sensors
connected to the IoT platform. These sensors are attached to critical machinery,
measuring various parameters like temperature, vibration, pressure, and power
consumption.

Collection of Data and Analysis:

The sensors continuously collect data in real time, and this data is transmitted to
the IoT platform. Advanced analytics algorithms process the data, identifying
patterns and anomalies. ML models are trained on historical data to analyze the
normal behavior of the equipment.

Predictive Analytics:
As the ML models are refined and improved over a period of time, they are cap­
able of predicting potential equipment failures. For instance, if abnormal vibra­
tion patterns are detected in a certain machine, it may indicate an impending
failure.

Alerts and Notifications:

The IoT platform is configured to generate alerts and notifications when a devi­
ation from the normal operating environment is identified. Maintenance teams
receive these alerts in real-time and provide insights into which equipment is at
risk of failure and the predicted time frame for the failure.

Proactive Maintenance:
Provided with predictive information, maintenance teams can schedule pro­
active maintenance activities. Instead of relying on fixed schedules or reactive
maintenance after a breakdown, they can suggest potential issues before they
escalate. This reduces downtime, increases the lifespan of equipment, and
minimizes the chances of unexpected failures.

CASE STUDY 2: PERSONALIZED HEALTH MONITORING

In the realm of personalized healthcare monitoring, a real-time case study
involves the integration of wearable devices and advanced data analytics to
tailor medical insights and interventions to individual patients. Consider a
Machine Learning in Internet of Things 79

scenario where a patient with chronic conditions, such as diabetes and hyper­
tension, is equipped with a wearable health monitor. This device continuously
collects real-time data on vital signs, activity levels, and glucose levels. The
data is seamlessly transmitted to a centralized healthcare platform utilizing
IoT technology. Advanced analytics algorithms on the platform process this
personalized health data, generating dynamic insights into the patient’s health
status. In the event of deviations from baseline parameters, the system triggers
automated alerts for healthcare providers and the patient, facilitating prompt
intervention. This proactive approach allows for timely medical adjustments
and empowers patients to actively manage their health, fostering a personalized
and preventative healthcare model.
Through personalized healthcare monitoring, patients benefit from
individualized treatment plans based on real-time data, ultimately improving
health outcomes and reducing the risk of complications. This case study
exemplifies the transformative potential of IoT-enabled personalized
healthcare, illustrating how the convergence of wearable devices and data
analytics can usher in a new era of patient-centric and data-driven healthcare
management.

4.6 FUTURE DIRECTIONS AND RESEARCH TRENDS

Looking ahead, this section explores the future and research tendencies in ML for
IoT. It discusses the concept of federated learning, where models are trained by sev­
eral IoT devices without centralized data storage, ensuring privacy and efficiency.
Additionally, it examines the role of edge computing in enabling decentralized intel­
ligence and reducing the dependence on cloud resources. The section focuses on
the importance of explainable AI in IoT, where ML models provide transparent and
interpretable outputs. Furthermore, it emphasizes the need for human-machine col­
laboration and interdisciplinary collaborations to manage the technical and societal
challenges posed by ML in IoT.

4.7 CONCLUSION
The chapter accomplishes by reviewing the key points discussed in this chapter. It
highlights the transformative impact of ML in enhancing intelligence and efficiency
within IoT systems.

REFERENCES
[1] Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2011). Edge computing: Vision and
challenges. IEEE Internet of Things Journal, 3(5), 637-646.
[2] Hastie, T., Tibshirani, R., & Friedman, J. (2009). The Elements of Statistical
Learning: Data Mining, Inference, and Prediction. Springer.
[3] Russell, S. J., & Norvig, P. (2009). Artificial Intelligence: A Modern Approach (3rd
ed.). Pearson.
80 IoT and Machine Learning for Smart Applications

[4] Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of Things
(IoT): A vision, architectural elements, and future directions. Future Generation
Computer Systems, 29(7), 1645-1660.
[5] Breiman, L. (1984). Classification and Regression Trees. CRC Press.
[6] Cortes, C., & Vapnik, V. (1995). Support-vector networks. ML, 20(3), 273-297.
[7] Liaw, A., & Wiener, M. (2002). Classification and regression by random forest.
R News, 2(3), 18-22.
 Role of Machine
5 Learning in Real-Life
Environment
Balasaraswathi V.R., Nathezhtha T.,
Vaissnavie V., and Rajashree

5.1 INTRODUCTION
Machine learning (ML) has a crucial role in leveraging technologies around Artificial
Intelligence (AI). ML is frequently referred to as AI due to its capacity for learning
and decision-making, but in truth, it is a subset of AI. Image Processing and Natural
language processing (NLP) are examples of machine learning approaches that can be
used to speed up data collection and format data. This can improve the ability to spot
clinical patterns and help with more accurate forecasts. Pattern recognition and data
analysis are a sort of data mining (DM) that enables computers to “learn” on their
own. Numerous advantages of ML include enhanced decision-making, accuracy, and
effectiveness. ML has emerged as a powerful tool in many fields, growing businesses,
and a wide range of applications [1]. Following are the categories of machine learning.

Supervised Learning: In supervised learning, the input is coupled with target labels,
and the system learns from labelled samples. The objective is to train the model to use
the labels provided to transfer the input data to the desired output. In order to forecast
or categorise previously unobserved data, the algorithm generalises from the labelled
data. Supervised learning algorithms encompass a variety of methods, such as Linear
Regression (LR), Logistic Regression, Decision Trees (DT), and SVM [2].

Unsupervised Learning: It is the process of finding patterns, structures, or

relationships in unlabelled data without using explicit target labels. The objective
is to gather relevant data points, minimise the complexity of the data, or draw
insightful conclusions. It is used for anomaly detection, dimensionality reduction,
and data clustering. Clustering techniques, such as k-means clustering and hierarch­
ical clustering, are employed to group similar data points based on their features.
Methods like principal component analysis (PCA) and t-distributed stochastic neigh­
bour embedding (t-SNE) are utilised to reduce the number of features while retaining
essential information. Anomaly detection techniques locate out-of-the-ordinary or
rare data points [3].

DOI: 10.1201/9781032623276-5 81
82 IoT and Machine Learning for Smart Applications

Reinforcement Learning: Through iterative experimentation, an agent acquires

the knowledge of interacting with the environment to optimise cumulative rewards.
The learning process involves the agent actively engaging with the environment and
receiving feedback in the form of rewards or punishments. The ultimate goal is to
discern optimal courses of action or policies that yield the highest cumulative reward
over time. Reinforcement learning has found successful applications in diverse areas
such as robotics, autonomous systems, and gaming. The agent, environment, actions,
rewards, and a policy that directs the agent’s decision-making process are crucial
elements in reinforcement learning [4].

5.2 MACHINE LEARNING IN HEALTHCARE

A variety of disease identification and treatments has been revolutionised by machine
learning, which has become a potent instrument in the medical field. ML algorithms can
analyse complicated patterns, generate precise predictions, and help healthcare practitioners
make educated decisions by utilising large volumes of patient data and advanced algorithms.
Healthcare with the help of ML is benefitted in many fields, as shown in Figure 5.1.

5.2.1 Use of Healthcare with ML

5.2.1.1 Improved Diagnosis
Large datasets including electronic health records, genetic data, and medical imaging
are used for analysis by machine learning algorithms. These algorithms are able to

Improved Diagnosis
• Learning patterns from medical data
• Assist healthcare professions on early diagnosis

Predictive Analytics
• Patient outcomes based on historical data
• Help healthcare providers allocate resources effectively

Drug Discovery and Development

• Analyzing large databases of molecular structures
• Expedites the discovery process

Streamlining Healthcare Operations

• Analyze medical records and extract relevant information
• Simplifying documentation

Fraud Detection and Cybersecurity

• Detect insurance fraud and billing anomalies
• Identify suspicious patterns

FIGURE 5.1 Role of ML in healthcare.

Role of Machine Learning in Real-Life Environment 83

recognise small patterns that would not be noticeable to human observers, resulting
in quicker and more accurate diagnoses. For instance, ML models in radiology can
discover anomalies in medical pictures or early indicators of diseases like cancer,
allowing for early intervention and better patient outcomes [5].
Liang et al. introduced an innovative classification framework known as Multi­
Level FC Fusion (MFC) to identify brain diseases. The model is developed utilising
both supervised and unsupervised learning techniques and methodical functional
Magnetic Resonance Imaging dataset tests. The effectiveness and generality of the
method are shown by the results, demonstrating strong classification performance
across various preprocessing pipelines and cross-validation schemes, which is a char­
acteristic feature of the proposed approach [6].

5.2.1.2 Personalised Treatment

Personalised treatment plans are one of machine learning’s main advantages in the
medical field. ML algorithms can forecast the best treatment options and dosages
by considering patient-specific aspects like medical history, genetic information,
and lifestyle data. This individualised strategy raises patient happiness, decreases
unfavourable reactions, and improves therapeutic effectiveness. In order to imple­
ment pre-emptive interventions and preventive care plans, machine learning models
can help identify patients who are more likely to experience the onset of particular
illnesses [7].
The Internet of Things, wearable technology, genetics, and image pro­
cessing marked the beginning of the digital era in medicine. AI technologies are
required to enable forecasts in order to personalise therapies. Concerns, including
explainability, liability, and privacy must be addressed for AI to be widely used in
healthcare [8].

5.2.1.3 Predictive Analytics

Based on past data, machine learning algorithms can forecast disease development
and patient outcomes. These algorithms may produce precise predictions about dis­
ease trajectories and identify high-risk individuals by examining a massive quantity of
patient data, including demographics, symptoms, test results, and therapy responses.
Predictive analytics can assist healthcare professionals in better resource allocation,
early detection of probable issues, and patient management techniques. Ahmed et al.
proposed a fuzzy model to diagnose the diabetic level with a membership value [9].
Many researchers focus on efficient prediction of cancer, which helps in the early
detection of disease and minimises the death rate [10].

5.2.1.4 Drug Discovery and Development

ML is also used in the field of finding new drugs and developing them [11]. ML
algorithms are capable of identifying promising medication candidates, predicting
their efficacy, and even optimising dosages by studying huge databases of molecular
structures, biological information, and clinical trial outcomes. This method speeds
up the research process, lowers expenses, and improves the likelihood that novel
treatments will be developed successfully.
84 IoT and Machine Learning for Smart Applications

Finding numerous novel pharmacological indications has been greatly aided

by chance. Drug-repositioning hypotheses could be created and validated with the
aid of algorithmic identification of patient-reported unexpected drug use in social
media [12].

5.2.1.5 Streamlining Healthcare Operations

Beyond diagnosis and treatment, ML is revolutionising administrative and oper­
ational activities in the healthcare industry. The documentation and coding
processes can be made simpler by using NLP algorithms to analyse medical
data and extract pertinent information. To increase operational effectiveness in
healthcare settings, machine learning models can also optimise resource alloca­
tion, forecast patient flow, streamline scheduling, and other techniques to detect
the patient’s stability [13].
By improving patient care overall and patient diagnosis, treatment, and care,
machine learning has the potential to completely transform the healthcare industry.
Machine learning can help healthcare providers make more accurate diagnoses, pro­
vide individualised treatment regimens, forecast disease outcomes, and streamline
operations by utilising the power of data and sophisticated algorithms. However,
ensuring the responsible and secure application of machine learning in healthcare
still depends on ethical considerations, privacy issues, and the requirement for human
oversight. As this discipline develops, it holds the potential to guide in a new era of
precision medicine and better patient outcomes across the board.
The research articles about machine learning in relation to healthcare are presented
in Table 5.1, which describes the diseases identified with its associated dataset and the
advantages of the proposed article. One of the widely used datasets for breast cancer
identification is Welikala et al. [14]. The Wisconsin Breast Cancer (Diagnosis) dataset
comprises 569 instances with 32 attributes, including an ID and a target variable.
Similarly, the Wisconsin Breast Cancer (Prognosis) dataset includes 198 instances
and 34 attributes, each with an ID and a target variable. Notably, the forecast dataset
underwent the removal of four instances with missing attribute values. The forecast
dataset exhibits a notable skew, with 151 instances of non-recurring outcomes and
47 instances of recurring outcomes. To address the challenge of imbalanced categor­
ization, two approaches, namely the algorithmic and data-centric approaches, were
employed in both the BC Wisconsin Diagnostic and Prognostic datasets. The first
tactic involves training the model to improve performance in minority classes by
employing cost-sensitive learning or applying a misclassification penalty. The results
[15,16] are depicted in Figure 5.2.

5.3 MACHINE LEARNING IN BUSINESS

Machine learning has spread across many facets of business, giving organisations
useful insights, automation capabilities, and enhanced decision-making procedures.
Some crucial business applications of ML are shown in Figure 5.3. Business processes
are assets of the organisation that are essential for providing value to the consumers
[20], and evolving technologies like ML and Deep Learning(DL) play a vital role in
the elevation of business.
 Role of Machine Learning in Real-Life Environment
TABLE 5.1
Survey of ML in Healthcare

References Disease ML Techniques Datasets Conclusion Advantages

[15] Cancer Logistic Regression. Wisconsin Importance early diagnoses experienced Early diagnosis can
KNN. SVM. RF. DT. Breast by both current and former breast reduce the risk of
and Naive Bayes (NB) Cancer cancer patients. death
classification Dataset
[16] Breast, lung, and Genetic Algorithm- Irvine Provides physicians with a diagnostic It acts as a general tool
cervical cancer Correlation Based repository tool by helping them to make a proper for extracting patterns
Datasets Feature Selection. diagnosis from several clinical
DT. SVM. Linear trials for various
Discriminant Analysis cancer diseases.
(LDA). NN
[17] Oral Potentially Random Forest Surgical Care Predict the risk of cancer for It is a unique model
Malignant Unit(SCU) Oral Potentially for accurate and
Disorders Dataset Malignant Disorders. Sensitivity: 0.82 affordable OPMDs/
(OPMD) Specificity: 0.9 Oral Squamous Cell
Carcinoma (OSCC)
screening.
Cost-effective tool for
predicting OPMDs.
[18] Oral Cavity Naive Bayes NB. Patients of 92 % Tumour level and the prediction of Accurate feature subset
Squamous-Cell Bagging of Naive University nodal status in patients with oral cavity is identified for
Carcinoma Bayes. KNN. J48. of Naples squamous cell carcinoma (OCSCC) and predicting the cancer
(OCSCC) boosting J48 oropharyngeal cancers is anticipated. in an accurate manner
[19] Coronavirus SVM Epidemiology 87% accurately categorise the patient's Accurate prediction
disease Dataset condition into no infection, mild
infection, and serious infection based
on the infection level.
86 IoT and Machine Learning for Smart Applications

0 20 40 60 80 100
Accuracy

FIGURE 5.2 (a) Accuracy of breast cancer prediction for Wisconsin dataset (b) accuracy of
cancer prediction Irvine dataset.

FIGURE 5.3 Role of machine learning in business.

Role of Machine Learning in Real-Life Environment 87

FIGURE 5.4 Accuracy of credit card fraud detection compared with the results.

Credit Card Fraud Detection

For computational intelligence algorithms, detecting fraud in credit card transactions
is a difficult problem that involves issues including idea drift, class imbalance, and
verification latency. Most learning algorithms include assumptions that are false
when used with fraud-detection systems (FDS) in the actual world. The fraud detec­
tion challenge is formalised in this work by outlining the operational parameters of
FDSs that process enormous volumes of credit card transactions [21].
For both cardholders and financial institutions, credit card fraud causes consider­
able financial loss. This study tries to identify high-class imbalanced data and changes
in the nature of fraud. Figure 5.4 shows the accuracy of credit card fraud detection
compared with the results. The European Card Benchmark dataset was used for com­
parative analysis, with ML methods to increase accuracy [22].
Healthcare systems are at risk from collusive fraud, which occurs when several
fraudsters work together to steal money from health insurance funds. Due to the simi­
larities to medical visits and the lack of labelled data, current approaches are facing
difficulty in detecting fraud. To identify suspicious patient behaviour, Fraud Auditor,
a three-stage visual analytics approach, integrates interactive co-visit networks,
enhanced community detection algorithms, and custom visualisations. The method’s
efficacy and usability were demonstrated through real-world examples and profes­
sional commentary [23]. Table 5.2 contains the widely used application for business
and the techniques used with its dataset.
 oo
oo

TABLE 5.2
Application for Business and the Techniques Used with Its Dataset

References Application ML Techniques Datasets Results Conclusion Advantages

[24] Credit Card LR. Credit Card Fraud LR: 99.23% Solves business Helps effectively to
Fraud RF. NB and Multilayer Detection RF:99.96%' problem. detect and prevent
Detection Perceptron (MLP) NB: 99.23%' frauds

loT and Machine Learning for Smart Applications

MLP:99.93%'
[25] Fake Image KNN. DT. Gaussian Naive In-house dataset Accuracy: 99% RF is comparatively Comparative
Detection Bayes. RF. Adaboost better results provide a
Classifier, and LR framework
[26] Fraud Detection Hybrid Technique - Credit Card Fraud Evaluated from Outliers highlights the Better results
Combination of Supervised Detection baseline to information about
and Un-Supervised learning Dataset. 5000 clusters risk
[27] Employee Chum CatBoost. RF. LR. DT. SVM Human Resource Accuracy: 95.4%' CatBoost Algorithm Predicts the risk
prediction and Information gives higher of chum in any
Retention Dataset (HRIS) prediction rate organisation
[28] Recommendation NB. SVM Movie Lens NA NB performs better Performs sentiment
systems Dataset analysis
Role of Machine Learning in Real-Life Environment 89

5.4 MACHINE LEARNING IN ROBOTICS AND AUTONOMOUS

SYSTEMS
Robotics and autonomous systems depend on machine learning to perceive, acquire,
and adapt to their surroundings. Robots can handle challenging situations and
make intelligent decisions using sensors and data analysis. Robots can benefit from
machine learning algorithms for tasks like object detection, path planning, and obs­
tacle avoidance. Robots can continuously enhance their performance and refine their
actions by learning from real-time data. ML is a key component of autonomous
vehicles’ ability to sense their environment, decipher traffic signs, and make quick
decisions. With the help of this technology, self-driving cars can manoeuvre safely
and effectively, lowering the risk of accidents [29-31]. Table 5.3 shows the survey
of ML in robotics and autonomous systems, Figure 5.5 (a) specifies the accuracy of
robotic-assisted urologic surgery, Figure 5.5 (b) shows the pedestrian detection in
an autonomous vehicle environment, Figure 5.5 (c) provides the analysis of crash
severity, and Figure 5.5 (d) shows the citrus fruit disease detection.
Machine learning algorithms are used to recognise and track specific objects or
features in the robot’s field of view [37]. This capability is essential for tasks such
as picking and placing objects, manipulating, or following a moving target. ML can
be used to optimise motion planning and control algorithms [38]. Reinforcement
learning (RL) algorithms can learn optimal control policies by trial and error to
perform tasks with more efficiency and adaptability. This is particularly useful in
applications like robot arm manipulation or mobile robot navigation. ML enables
robots to interact with humans in natural and intuitive ways. NLP techniques can
be used to understand and respond to human speech, while sentiment analysis can
help robots gauge user emotions. This enhances the robot’s ability to collaborate
and communicate effectively with humans. Machine learning algorithms are instru­
mental in autonomous navigation for robots. By analysing sensor data and learning
from past experiences, it can navigate complex environments, avoid obstacles, and
plan efficient paths. This is crucial for applications such as autonomous drones,
self-driving cars, or warehouse robots. Figure 5.6 shows the different applications
of ML in Robotics.
Machine learning techniques can optimise task planning and scheduling for robots.
By analysing historical data, robots can learn to estimate task durations, allocate
resources efficiently, and improve overall task execution. Machine learning models
can analyse environmental data, including soil moisture levels, weather conditions,
and crop requirements, to optimise irrigation scheduling. By learning from historical
data, these algorithms can make predictions and adapt irrigation practices to conserve
water and improve crop health. The survey of ML in NLP is shown in Table 5.4.
NLP enables computers to comprehend, examine, and produce human language
and depends heavily on machine learning. Using ML approaches for sentiment ana­
lysis, text data can be classified as having positive, negative, or neutral feelings. By
training on labelled data, these algorithms can accurately determine the sentiment
expressed in customer reviews, social media posts, or other text sources. Businesses
can utilise this to gain deeper insight into the attitudes and viewpoints of a sizable
client base.
 <0
o
TABLE 5.3
Survey of ML in Robotics and Autonomous Systems

References Application ML Techniques Datasets Results Conclusion Advantages

[32] Robotic-assisted Combination of ML and Johns Hopkins The accuracy. ML-guided Surgeon deals
urologic surgery Augmented Reality University - sensitivity, and autonomous the novel data
(AR) Intelligent Systems specificity metrics surgery
Institute (JHU- achieve rates of
ISI) Gesture and 92.5%. 95.8%.
Skill Assessment and 88.8%'.
Working Set respectively.

loT and Machine Learning for Smart Applications

(JIGSAWS)
[33] Pedestrians KNN. SVM and NB Sample points Accuracy Provides more Scalability of
detection in Classifier generated by 3D KNN:90.72%' accurate results at ML enables
autonomous Light Detection and SVM Linear: 96.29 all type lighting the detection
vehicle Ranging (LIDAR) SVM conditions of dissimilar
environment Quadratic: 89.28%' objects
[34] Analyses the crash Classification and Crash reports received Accuracy XGBoost model Reduces the
severity regression tree (CART). in California XGBoost:64.10%' recognises the severity of
SVM. extreme Gradient from Jan- 2019 to CART:66.67 % injured crashes in AV-involved
Boosting (XGBoost) Oct- 2020 SVM:71.79 % an efficient manner crashes
[35] Detects the citrus K Means clustering. Citrus fruit dataset Accuracy Detects the citrus Powerful tool
fruit disease Artificial Neural ANN:88.96%' infection in less for automated
automatically Network(ANN) and SVM: 93.12%' computational detection
SVM effort
[36] Automation of SVM and Support Vector Agriculture Crop - Economical approach Computationally
Farm Irrigation Regression( S VR) Production in India for automation in efficient
System Dataset agriculture approach
Role of Machine Learning in Real-Life Environment 91

FIGURE 5.5 Accuracy of (a) robotic assisted urologic surgery (b) pedestrians detection
in autonomous vehicle environment (c) analysis of crash severity (d) citrus fruit disease
detection.

FIGURE 5.6 Applications of ML in robotics.

 KJ

TABLE 5.4
Survey of ML in NLP

References Application ML Techniques Datasets Results Conclusion Advantages

[39] Sentiment analysis Unsupervised Airport and - Analyses the online Helps to develop
using online learning Airline reviews, reviews in the the managerial
reviews tourist reviews, domain of tourism strategies for
etc. and hospitality. consumers.
[40] Determines the Social media
Binary Classification Twitter- ML models identify Negative opinion

loT and Machine Learning for Smart Applications

users view in an platform like Positive Opinion:61.66% the sentiments helps to identify
automated way twitter and Negative Opinion:38.34% in the data and the key fears and
reddit Reddit also determine reservation.
Positive opinion: 71.72% the level of
Negative opinion: 28.28%' acceptance.
[41] Automatically DT. Bernoulli NB. Movie Review Accuracy MNB provides Helps to realise the
symbolises the SVM. Maximum Dataset BNB: %' better result expression of
expression of Entropy (ME), and DT: 80.17%' feelings around
feelings for Multinomial NB. SVM:87.33%' everything like
movie. ME: 60.67%' product, social
MNB: 88.50%' media, etc.
[42] Analyse and MNB. Bernoulli NB. Al-Khaleej dataset SVM performs better Helps to maintain Useful for multiple
categorises the Stochastic Gradient the significant domains, mostly on
Arabic text Descent (SGD). information in social media.
LR. SVM classifier several domains.
Role of Machine Learning in Real-Life Environment 93

Machine learning models can find and classify named entities within a text, such
as people, organisations, places, and dates. Named Entity Recognition (NER) can be
useful for applications, including information extraction, search engines, and recom­
mendation systems. Machine learning algorithms are trained to accurately recognise
and extract named things from annotated data.
Text documents are categorised into predetermined groups by machine learning
algorithms for purposes including spam detection, subject categorisation, sentiment
analysis, and content filtering.
Machine translation systems improve with machine learning approaches like
neural machine translation (NMT), enhancing accuracy by recognising contextual
cues and grammatical patterns in multiple languages.
Question-answering systems use machine learning techniques to analyse and pro­
vide answers to questions using text matching, natural language comprehension,
and information retrieval, enabling them to understand queries and extract relevant
information.

REFERENCES
1. S. V. Mahadevkar et al., “A Review on Machine Learning Styles in Computer Vision—
Techniques and Future Directions,” IEEE Access, vol. 10, pp. 107293-107329, 2022,
DOI: 10.1109/ACCESS.2022.3209825.
2. J. Camacho, G. Macia-Fernandez, N. M. Fuentes-Garda and E. Saccenti, “Semi­
Supervised Multivariate Statistical Network Monitoring for Learning Security
Threats,” IEEE Transactions on Information Forensics and Security, vol. 14, no. 8,
pp. 2179-2189, Aug. 2019, DOI: 10.1109/TIFS.2019.2894358.
3. S. J. Pan and Q. Yang, “A Survey on Transfer Learning,” IEEE Transactions on
Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345-1359, Oct. 2010,
DOI: 10.1109/TKDE.2009.191.
4. X. Ma and W. Shi, “AESMOTE: Adversarial Reinforcement Learning with
SMOTE for Anomaly Detection,” IEEE Transactions on Network Science and
Engineering, vol. 8, no. 2, pp. 943-956, 1 Apr.-June 2021, DOI: 10.1109/
TNSE.2020.3004312.
5. M. Li and Z.-H. Zhou, “Improve Computer-Aided Diagnosis with Machine Learning
Techniques Using Undiagnosed Samples,” IEEE Transactions on Systems, Man, and
Cybernetics - Part A: Systems and Humans, vol. 37, no. 6, pp. 1088-1098, Nov. 2007,
DOI: 10.1109/TSMCA.2007.904745.
6. Y. Liang and G. Xu, “Multi-Level Functional Connectivity Fusion Classification
Framework for Brain Disease Diagnosis,” IEEE Journal of Biomedical and
Health Informatics, vol. 26, no. 6, pp. 2714-2725, June 2022, DOI: 10.1109/
JBHI.2022.3159031.
7. K. Paranjape, M. Schinkel and P. Nanayakkara, “Short Keynote Paper: Mainstreaming
Personalized Healthcare - Transforming Healthcare through New Era of Artificial
Intelligence,” IEEE Journal of Biomedical and Health Informatics, vol. 24, no. 7,
pp. 1860-1863, July 2020, DOI: 10.1109/JBHI.2020.2970807.
8. S. Zhang, S. M. H. Bamakan, Q. Qu and S. Li, “Learning for Personalized
Medicine: A Comprehensive Review from a Deep Learning Perspective,” IEEE
Reviews in Biomedical Engineering, vol. 12, pp. 194-208, 2019, DOI: 10.1109/
RBME.2018.2864254.
94 IoT and Machine Learning for Smart Applications

9. U. Ahmed, et al., “Prediction of Diabetes Empowered with Fused Machine

Learning,” IEEE Access, vol. 10, pp. 8529-8538, 2022, DOI: 10.1109/
ACCESS.2022.3142097.
10. N. Fatima, L. Liu, S. Hong and H. Ahmed, “Prediction of Breast Cancer, Comparative
Review of Machine Learning Techniques, and Their Analysis,” IEEE Access, vol. 8,
pp. 150360-150376, 2020, DOI: 10.1109/ACCESS.2020.3016715.
11. N. R. C. Monteiro, B. Ribeiro and J. P. Arrais, “Drug-Target Interaction Prediction: End-
to-End Deep Learning Approach,” IEEE/ACMTransactions on Computational Biology
and Bioinformatics, vol. 18, no. 6, pp. 2364-2374, 1 Nov.-Dec. 2021, DOI: 10.1109/
TCBB.2020.2977335.
12. B. Ru, D. Li, Y. Hu and L. Yao, “Serendipity—A Machine-Learning Application
for Mining Serendipitous Drug Usage from Social Media,” IEEE Transactions
on NanoBioscience, vol. 18, no. 3, pp. 324-334, July 2019, DOI: 10.1109/
TNB.2019.2909094.
13. B.-S. Lin, et al., “Fall Detection System withArtificial Intelligence-Based Edge Computing,”
IEEE Access, vol. 10, pp. 4328-4339, 2022, DOI: 10.1109/ACCESS.2021.3140164.
14. R. A. Welikala, et al., “Automated Detection and Classification of Oral Lesions Using
Deep Learning for Early Detection of Oral Cancer,” IEEE Access, vol. 8, pp. 132677­
132693, 2020, DOI: 10.1109/ACCESS.2020.3010180.
15. Ak, Muhammet Fatih, “A Comparative Analysis of Breast Cancer Detection and
Diagnosis Using Data Visualization and Machine Learning Applications,” Healthcare,
vol. 8., no. 2, pp. 1-23, 2020.
16. Hsu, Ching-Hsien, et al., “Effective Multiple Cancer Disease Diagnosis Frameworks
for Improved Healthcare Using Machine Learning,” Measurement, vol. 175, pp. 109­
145, 2021.
17. Wang, Xiangjian, et al., “A Personalized Computational Model Predicts Cancer Risk
Level of Oral Potentially Malignant Disorders and Its Web Application for Promotion
of Non-invasive Screening,” Journal of Oral Pathology & Medicine, vol. 49, no. 5, pp.
417-426, 2020.
18. Romeo, Valeria, et al., “Prediction of Tumor Grade and Nodal Status in Oropharyngeal
and Oral Cavity Squamous-Cell Carcinoma Using a Radiomic Approach,” Anticancer
Research, vol. 40, no. 1, pp. 271-280, 2020.
19. Guhathakurata, Soham, et al., “A Novel Approach to Predict COVID-19 Using Support
Vector Machine,” Data Science for COVID- 19, vol. 1, pp. 351-364, 2021.
20. G. Polancic, S. Jagecic and K. Kous, “An Empirical Investigation of the Effectiveness
of Optical Recognition of Hand-Drawn Business Process Elements by Applying
Machine Learning,” IEEE Access, vol. 8, pp. 206118-206131, 2020, DOI: 10.1109/
ACCESS.2020.3034603.
21. A. Dal Pozzolo, G. Boracchi, O. Caelen, C. Alippi and G. Bontempi, “Credit Card Fraud
Detection: A Realistic Modeling and a Novel Learning Strategy,” IEEE Transactions
on Neural Networks and Learning Systems, vol. 29, no. 8, pp. 3784-3797, Aug. 2018,
DOI: 10.1109/TNNLS.2017.2736643.
22. F. K. Alarfaj, I. Malik, H. U. Khan, N. Almusallam, M. Ramzan and M. Ahmed,
“Credit Card Fraud Detection Using State-of-the-Art Machine Learning and Deep
Learning Algorithms,” IEEE Access, vol. 10, pp. 39700-39715, 2022, DOI: 10.1109/
ACCESS.2022.3166891.
23. J. Zhou, et al., “FraudAuditor: A Visual Analytics Approach for Collusive Fraud in
Health Insurance,” IEEE Transactions on Visualization and Computer Graphics, vol.
29, no. 6, pp. 2849-2861, 1 June 2023, DOI: 10.1109/TVCG.2023.3261910.
Role of Machine Learning in Real-Life Environment 95

24. Armedja, Dejan, et al., “Credit card fraud detection-machine learning methods,”
International symposium INFOTEH-JAHORINA (INFOTEH), 2019.
25. Hamid, Yasir, et al., “An improvised CNN model for fake image detection,”
International Journal of Information Technology, vol. 15, no. 1, pp. 5-15, 2023.
26. Carcillo, Fabrizio, et al,. “Combining unsupervised and supervised learning in credit
card fraud detection,” Information Sciences, vol. 557, pp. 317-331, 2021.
27. Jain, Nishant, Abhinav Tomar and Prasanta K. Jana, “A Novel Scheme for Employee
Churn Problem Using Multi-Attribute Decision Making Approach and Machine
Learning,” Journal of Intelligent Information Systems, vol. 56, pp. 279-302, 2021.
28. Marappan, Raja and S. Bhaskaran, “Movie Recommendation System Modeling Using
Machine Learning,” International Journal of Mathematical, Engineering, Biological
and Applied Computing, pp. 12-16, 2022.
29. G. Jia, H.-K. Lam, S. Ma, Z. Yang, Y. Xu and B. Xiao, “Classification of
Electromyographic Hand Gesture Signals Using Modified Fuzzy C-Means Clustering
and Two-Step Machine Learning Approach,” IEEE Transactions on Neural Systems and
Rehabilitation Engineering, vol. 28, no. 6, pp. 1428-1435, June 2020, DOI: 10.1109/
TNSRE.2020.2986884.
30. A. A. Alhussan, D. S. Khafaga, E.-S. M. El-Kenawy, A. Ibrahim, M. M. Eid and A. A.
Abdelhamid, “Pothole and Plain Road Classification Using Adaptive Mutation Dipper
Throated Optimization and Transfer Learning for Self Driving Cars,” IEEE Access,
vol. 10, pp. 84188-84211, 2022, DOI: 10.1109/ACCESS.2022.3196660.
31. A. Altameem, A. Kumar, R. C. Poonia, S. Kumar and A. K. J. Saudagar,
“Early Identification and Detection of Driver Drowsiness by Hybrid Machine
Learning,” IEEE Access, vol. 9, pp. 162805-162819, 2021, DOI: 10.1109/
ACCESS.2021.3131601.
32. Ma, Runzhuo, et al., “Machine Learning in the Optimization of Robotics in the
Operative Field,” Current Opinion in Urology, vol. 30, no. 6, p. 808, 2020.
33. Navarro, Pedro J., et al., “A Machine Learning Approach to Pedestrian Detection for
Autonomous Vehicles Using High-Definition 3D Range Data,” Sensors, vol. 17, no. 1,
pp. 1-15, 2016.
34. Chen, Hengrui, et al., “Exploring the Mechanism of Crashes with Autonomous
Vehicles Using Machine Learning,” Mathematical Problems in Engineering, vol. 1,
pp. 1-10, 2021.
35. Doh, Benjamin, et al., “Automatic Citrus Fruit Disease Detection by Phenotyping
Using Machine Learning,” 25th International Conference on Automation and
Computing (ICAC), pp. 1-5, 2019.
36. Vij, Anneketh, et al., “IoT and Machine Learning Approaches for Automation of Farm
Irrigation System,” Procedia Computer Science, vol. 1, pp. 1250-1257, 2020.
37. S. A. Pedram, P. W. Ferguson, M. J. Gerber, C. Shin, J.-P. Hubschman and J. Rosen,
“A Novel Tissue Identification Framework in Cataract Surgery Using an Integrated
Bioimpedance-Based Probe and Machine Learning Algorithms,” IEEE Transactions
on Biomedical Engineering, vol. 69, no. 2, pp. 910-920, Feb. 2022, DOI: 10.1109/
TBME.2021.3109246.
38. D.-P. Tran, G.-N. Nguyen and V.-D. Hoang, “Hyperparameter Optimization
for Improving Recognition Efficiency of an Adaptive Learning
System,” IEEE Access, vol. 8, pp. 160569-160580, 2020, DOI: 10.1109/
ACCESS.2020.3020930.
39. Jain, Praphula Kumar, Rajendra Pamula and Gautam Srivastava, “A Systematic
Literature Review on Machine Learning Applications for Consumer Sentiment
Analysis Using Online Reviews,” Computer Science Review, vol. 41, pp. 100413, 2021.
96 IoT and Machine Learning for Smart Applications

40. Bakalos, Nikolaos, Nikolaos Papadakis and Antonios Litke, “Public Perception of
Autonomous Mobility Using ML-Based Sentiment Analysis Over Social Media Data,”
Logistics, vol. 4, no. 2, p. 12, 2020.
41. Muaad, Abdullah Y., et al., “An Effective Approach for Arabic Document Classification
Using Machine Learning,” Global Transitions Proceedings, vol. 3, no. 1, pp. 267­
271, 2022.
42. Rahman, Atiqur and Md Sharif Hossen, “Sentiment Analysis on Movie Review Data
Using Machine Learning Approach,” International Conference on Bangla Speech and
Language Processing (ICBSLP), pp. 1-4, 2019.
 Efficient Blockchain­
6 Based Edge Computing
System Using Transfer
Learning Model
P. Sivakumar, S. Nagendra Prabhu,
S.K. Somasundaram, J. Uma Maheswari,
and Murali Murugan

6.1 INTRODUCTION
The IoT devices are integrated in several applications like gesture-tracking devices
with a high power and capability. The IoT system has consumed a maximum power
because of its broad usage. Many conventional IoT-based cloud communication are
performed and redesigned by using a mobile edge computing (MEC) system to reduce
computing power. The MEC system is integrated among sensor medium IoT nodes
and the cloud environment [1]. Thus, the MEC system achieved a lower computation
power using it but the security of the system is required. Decentralized technology
like blockchain is integrated with an MEC to achieve anonymity, secrecy, and privacy
of IoT cloud data.
Blockchain technology is based on bitcoin and has characteristics of decentraliza­
tion, transparency, and openness. Generally, this technology has an addition of string
data to a blockchain by encryption [2]. This technology is implemented in many con­
fidential sectors like banking, registration, and military data using a trust approach
based on credit identification. The blockchain can solve the third-party issue by dir­
ectly connecting two strange entities. It was a distributed system through consensus
mechanisms and node verification [3]. Thus, the distributed blockchain architecture
of IoT is presented for a value transfer by transmitting data.
This chapter implements the blockchain-based MEC (BMEC) for IoT data [4]. To
achieve a higher security and scheduling strategy, the TL-based clone block identi­
fication (CBI) is performed for the detection of clone block attacks. Therefore, the
proposed system carried a BMEC with a TL-based CBI to achieve an efficient system.
The proposed model’s performance is compared with a conventional method. The
rest of the work carried a related work in Section 6.2 and the preliminary of TL-based
MEC in Section 6.3. Section 6.4 presents the proposed BMEC system, and Section

DOI: 10.1201/9781032623276-6 97
98 IoT and Machine Learning for Smart Applications

6.5 includes experimental results. Finally, Section 6.6 includes a conclusion followed
by references.

6.2 RELATED WORKS

Samy et al. [4] presented a blockchain system for task offloading security in MEC
systems. This method achieved a better performance with better delay and energy
consumption. Next, the unmanned aerial vehicle (UAV) system provided secure
aerial computing architecture using blockchain, which was mentioned in [5]. This
system has ensured the security of computation offloading among mobile users and
UAVs. The paper [6] presented by W. Wang et al. that implemented a triple real­
time trajectory privacy protection methodology in MEC. W. Wu et al. [7] presented
a new approach called the Practical Byzantine Fault Tolerance (PBFT) method for
overcoming a committee selection issues and to improve network’s security. Next,
X. Wang et al. [8] developed a Deep Reinforcement Learning (DRL) based method
using a Mean Field Theory (MFT).
In [9], a blockchain-based deep actor-critic method for task offloading is
implemented. This method is used to secure the system and solve computation
offloading issues. Next, the blockchain-based conditional privacy-preserving authen­
tication method is presented in [10]. This system implicated permission for vehicular
edge computing. S. Ma et al [11] developed a Byzantine Fault Tolerance (BFT) model
to improve the performance. The deterministic network calculus approach is used to
analyze the BFT and HotStuff consensus methods.
In [12] a Service Providers location using a blockchain network is presented. This
system has various propagation delays when performing computations. P. Kumar
et al. [13] developed a mechanism called SAE-ABIGRU to detect Intrusion. This
method is used to predict the MEC link load behaviors of the Road Side Unit (RSU)
server.
J. Liu et al. [14] proposed a scheme for edge-based networks to increase security
and efficiency. They introduce multiple attribute schemes to support a large attribute
universe and alleviate management burdens. Edge computing is utilized to speedily
respond to vehicle requests and assist resource-constrained computation. Also, to
assure the security of users, a cooperative key generation has been used.
X. Huang et al. [15] present a hierarchical federated learning network enabled by
a multi-layer blockchain. This allows minimum latency without compromising the
security level of the network. Further, the bottleneck analysis has been carried out
with imbalanced data distribution.
P. Liu et al. [16] proposed a P2P consortium blockchain-based trading system
for resources to incentivize more participants in the context of security and privacy
concerns. They focus on the pricing strategy and user concerns, ensuring trust,
security on transactions, and privacy of a user without a central authority. C. Xu et al.
[1] focus on privacy computing technology and develop a secure architecture based
on blockchain.
S. Fugkeaw et al. [17] suggest an IoT aggregation and data transmission method
using digital signing and encryption, and they also highlight the collaboration between
fog nodes and blockchain. They introduce an encryption method combined with a
Efficient Blockchain-Based Edge Computing System 99

privacy-preserving access policy and a lightweight policy update algorithm for the
effective management of data. Comparative analysis and experiments demonstrate
the scheme’s computation cost and performance. The hybrid system is proposed by
Y. Zhong et al. [18] to design a double-layer mining service offloading. The proposed
system combines MEC with a blockchain. The task of offloading is handled by the
algorithm of actor-critic (A3C) algorithm and double auction.
Ren et al. [19] have introduced a task-offloading strategy known as blockchain­
based trust-aware task offloading strategy, which leverages blockchain technology
and trust-awareness. The authors employ smart contracts to facilitate secure and
automated task offloading. To mitigate network insider attacks, they incorporate
a recommendation filtering technique and a trust penalty measure. Furthermore,
the researchers analyze the challenges related to time efficiency when integrating
blockchain into their proposed system.
In [20], a blockchain authentication scheme is proposed for securing the messages.
The authors used edge servers and a lightweight message authentication algo­
rithm, reducing the computational overhead. Du et al. [21] present a system called
Blockchain-based Intelligent Edge Cooperation System (BIECS), which combines
blockchain and intelligent edge cooperation to enable efficient resource sharing.
Tian et al. [22] introduce Blockchain-based Secure Searching for Metadata
(BSSMeta), a secure searching approach for metadata in MEC using blockchain
technology. The authors employ various techniques such as lightweight proxy re­
encryption, master-slave smart contracts, and buffer uploading. These measures
enable efficient searching and uploading of metadata within the system. Chavhan
et al. [23] present a security scheme for vehicle platooning that integrates edge com­
puting and blockchain technology. The authors designed a security architecture and
implemented smart contracts that were seamlessly integrated with the Simulation of
Urban Mobility Traffic Control Interface Application Programming Interface. This
integration ensures the establishment of a secure environment for vehicle platooning.
L. Qi et al. [24] present a cooperative proof-of-work (PoW) strategy named
Relay-PoW, which reduces the consumption of energy and improves the efficiency of
resource utilization. They proposed a supervision group mechanism and parallel relay
mining to increase the throughput and ensure security. B. Jiang et al. [25] propose an
attribute-based encryption technique, blockchain, and smart contracts for access con­
trol and event traceability.
Lang et al. [26] introduce a framework for cooperative carbon monoxide (CO)
detection and utilize blockchain technology for secure handover in Vehicle Edge
Computing. The authors incorporated a consensus mechanism to develop models
for vehicle mobility and cooperative handover to guarantee the immutability of
offloaded data. J. Li et al. [27] propose a blockchain-assisted access control technique
for healthcare systems in fog computing. They utilize smart contracts and a multi­
authority attribute-based encryption scheme to ensure data transparency, reliability,
and confidentiality.
Y. Li et al. [28] propose a blockchain-based data integrity verification scheme
for cloud-edge computing environments. They propose a detailed scheme without
a trusted third party, consisting of data upload and verification phases. R. Li et al.
[29] propose a secure blockchain-based access control framework for the Industrial
100 IoT and Machine Learning for Smart Applications

Internet of Things (IIoT). They design an access frequency-based network sharding

scheme to improve scalability and a Bloom filters-based privacy protection scheme.
X. Ding et al. [30] consider the case where IoT devices require assistance from
edge servers for data analysis and storage-related tasks, which affect their profits
based on the resources purchased. They propose an allocation strategy where IoT
devices can purchase resources from various edge servers, and they can allocate
a limited budget to purchase the resources to increase the profit. In their study,
L. Sun et al. [31] address the heterogeneous and dynamic nature of fog nodes (FNs)
in fog computing resource management. They propose a truthful auction for the
fog system (TAFS) to incentivize FNs to satisfy application demands with guaran­
teed performance. It considers both delay tolerance and resource availability for all
applications.
Du et al. [32] introduce a matching mechanism based on smart contracts aimed
at establishing renting associations between Distribution System Operators (DSOs)
and Energy Consumption Nodes (ECNs) with the goal of maximizing social welfare.
Additionally, the authors introduce a double auction technique to calculate the pricing
for the winning applicants.
In their work, Nguyen et al. [33] propose a proof-of-reputation consensus mech­
anism based on a block verification strategy. Their research focuses on developing
a multi-objective function within the BMEC system. This function aims to maxi­
mize the system utility by optimizing various aspects, including offloading decisions,
resource, and power allotments.

6.3 PRELIMINARIES
In this section, the TL-based MEC is presented for an IoT application. This TL model
is based on the Convolutional Neural Network (CNN) structure of the DL approach.
The TL model is applied for fast computing, and it allocates a minimum amount of
time to learn the huge dataset. This TL model achieved a robustness of the system and
also computational delay is reduced in it due to its shortcut.

TL Model structure
In this MEC system, the VGG16-TL model is used for IoT data computing. The
VGG16 model provided a better system performance with a higher accuracy but does
not reduce the learning device.

VGG16 Description
The VGG16 is applied to a large-scale dataset that provides an ImageNet. This model
reduced the delay of learning time and also minimized insufficient data. Figure 6.1
shows the VGG16 block diagram that consists of a total of 16 layers, including 13
convolutional layers and 3 dense layers.
From Figure 6.1, the VGG16 consisted of convolutional and dense layers. The
dataset is used to learn the dense layer weights. Several convolutional layers are
categorized, namely 64 filters in 2 layers, 128 filters in 2 layers, 256 filters in 3 layers,
and 512 filters in 3 layers. All the convolutional filters provided a 3 x 3 size with a
Efficient Blockchain-Based Edge Computing System 101

FIGURE 6.1 VGG16 Transfer-learning models.

FIGURE 6.2 Architecture of TL-based MEC in IOT.

maximum pooling layer of 2 x 2 size. This VGG16 has two dense layers such as one
as a hidden layer with 128 neurons and another as an output layer. There are approxi­
mately 138 million parameters provided for the network.

TL-Based MEC in IOT

The VGG16 TL model is applied to IoT data in the MEC system, as shown in
Figure 6.2. The datasets are learned by using a Graphics Processing Unit (GPU)-
enabled cloud in this system. The trained dataset is processed by the Edge server
and provided with image data using visual sensors. The fine-tunings are performed
with a small amount of ground-labeled data in this system. Next, the edge server
processes edge computing using an installed sensor. The activity of an edge server
is recognized using a VGG16 classifier. The predicted inference is sent to the cloud
server. Then, a cloud server takes appropriate action based on present inference and
stored information.
102 IoT and Machine Learning for Smart Applications

FIGURE 6.3 Proposed architecture of BMEC system.

6.4 PROPOSED SYSTEM

The VGG16 TL method was efficient for a MEC-IoT system due to its minimum
delay in scheduling time. However, the security of an IoT data in the MEC system is
its biggest threat. To obtain security, the blockchain technology is integrated to TL-
based MEC system. Therefore, the proposed architecture of the MEC system based
on the blockchain is shown in Figure 6.3. This system comprised a mobile terminal
device, cloud servers, edge servers, and blockchain.
The IoT provides a mobile terminal device that is integrated with a block with
smart contracts. In this layer, the BMEC generated a blockchain account. The edge
servers are interconnected in the edge network layer. These layers have specific
blockchains on the edge layer and are also used to monitor the BMEC transaction.
Whenever blockchain transactions are processed, new blocks are created. The cloud
servers are interconnected and provide a blockchain network. The proposed system
can achieve a speedy transaction in the cloud computing layer by optimizing the
blockchain topology.

TL-Based Clone Block Identification

To enhance security, clone block attack detection is the major concern in the BMEC
system. The clone attack of the block was an illegal attacker. This attack imitates the
authorized block and gathers its legal information and copied to another blocks. The
original ID and hash key verification data are cloned in various locations in the net­
work. Figure 6.4 shows the clone block attack architecture that consists of an edge
node, a legal block, and three clone blocks. The clone blocks I, II, and III have fake
information about the original block. The fake content in clone block I is greater
than clone block 2 and clone block 3. Clone block II fake content is larger than clone
block III.
Efficient Blockchain-Based Edge Computing System 103

FIGURE 6.4 Clone blocks attack.

To identify this issue, the CBI is performed using a VGG16 TL model shown in
Figure 6.4. The TL model is very effective in prediction by training a few data sets.
The clone blocks are identified with their higher classification and prediction strategy.
This proved that the proposed TL-based CBI can be used to improve security in the
BMEC system. Therefore, the proposed BMEC system achieved both scheduling
accuracy and security using TL-based CBI.

VGG Tuning
The parameters in VGG16 include the weights and biases of these layers, which are
learned through the training process. These parameters define the network’s ability to
extract features from input data and make accurate predictions.
In deep learning, tuning of parameters is an essential step for the performance
optimization of a model. It involves adjusting the values of the model’s parameters
to achieve better accuracy or other desired metrics. Grey Wolf Optimization is a
metaheuristic algorithm based on the social behavior of wolves. It mimics the hunting
behavior of gray wolves to find the optimal solutions.
The objective function could be a combination of different metrics that capture
the system’s performance, such as classification accuracy, system delay, and security
measures. For example, you could define the objective function as a weighted sum of
accuracy and delay, where accuracy represents the model’s performance in correctly
classifying data, and delay represents the time taken for processing.
The Grey Wolf Optimization (GWO) algorithm would then be utilized to search the
parameter space of VGG16 and find the optimal values that maximize the objective
function. GWO is a population-based algorithm that emulates the hunting behavior of
104 IoT and Machine Learning for Smart Applications

gray wolves, consisting of alpha, delta, beta, and omega wolves representing different
positions in the search space.
During the optimization process, each wolf in the GWO algorithm corresponds
to a set of parameters in VGG16. The position of a wolf represents a potential solu­
tion in the parameter space. By iteratively updating the positions of wolves based on
their hunting behavior, GWO explores the parameter space and gradually converges
toward the optimal solution.
The behavior of wolves in GWO involves three main steps: encircling, attacking, and
updating positions. These steps can be related to the optimization process as follows:

Encircling: In this step, the GWO algorithm performs a search around the current
positions to explore new potential solutions. This simulates the exploration of
different parameter combinations in VGG16.

Attacking: The GWO algorithm identifies the most promising positions and updates
them to move closer to the optimal solution. This step corresponds to refining the par­
ameter values of VGG16.

Updating positions: The GWO algorithm changes the positions of wolves based on
their hunting behavior and the positions of the alpha, beta, delta, and omega wolves.
This step helps the algorithm converge toward the optimal solution and fine-tune the
parameters of VGG16.
Applying GWO-based tuning to VGG16, the objective function defined earlier
guides the optimization process to find the parameter values that maximize the desired
goals of efficient security and lower delay in the BMEC system. The GWO algorithm
explores the parameter space of VGG16, updating the weights and biases, until an
optimal solution is found. The pseudocode for the proposed GWO-based parameter
tuning is given below:

• Initialize a population of wolves (alpha, beta, delta, omega)

• Initialize VGG16 with random parameter values
• Define the objective function
• Function objective_function (parameters)
• Load VGG16 with the given parameters
• Train VGG16 on a labeled dataset
• Evaluate VGG16’s performance on a validation set
• Compute the objective function value based on the desired metrics
• Return objective_value
• Initialize the maximum number of iterations
• Initialize convergence criteria
• While not converged and iterations < max_iterations
• For each wolf in the population
• Encode the parameters of the wolf into VGG16
• Compute the objective function value using objective_function()
• Sort the population based on the objective function values
Efficient Blockchain-Based Edge Computing System 105

• Update positions of the wolves

• For each wolf in the population
• Update the positions based on the hunting behavior of wolves
• Perform parameter adjustments
• For each wolf in the population
• Adjust the parameters in VGG16 based on the positions of the wolves
• Increment the iteration counter
• Select the best wolf with the highest objective function value
• Retrieve the parameters from the best wolf
• Return the tuned parameters of VGG16

Through this iterative optimization process, the tuned VGG16 model achieves
improved performance, leading to better scheduling strategies, efficient security, and
reduced delay in the BMEC-IoT system compared to conventional methods. The
transfer learning (TL) based CBI further leverages the optimized VGG16 model to
improve the overall efficiency and effectiveness. The GWO algorithm explores the
parameter space and refines the parameters based on the hunting behavior of wolves.
The tuned VGG16 model contributes to the improved performance of the BMEC-IoT
system, resulting in better scheduling.

6.5 EXPERIMENTAL RESULTS

This section presented an experimental result of the proposed BMEC system. At first,
consider ten blocks for blockchain sharing. Assign the connection weight among
every block randomly. In this method, the TL-based CBI is used to recognize the
next block generation. From Figure 6.5, the blocks in the edge server are directly

FIGURE 6.5 Total delay in proposed BMEC system.

106 IoT and Machine Learning for Smart Applications

TABLE 6.1
Performance Table of Scheduling Accuracy

Methods Scheduling Accuracy

TL-based MEC 98.76

Proposed BMEC 99.54

FIGURE 6.6 Performance of scheduling accuracy.

proportional to the total delay in BMEC, where both the parameters increase non-
linearly. This showed that the system can enhance the throughput transaction and
have low latency of computation.
Table 6.1 and Figure 6.6 show the proposed BMEC and TL-based MEC
performances in scheduling accuracy. The proposed method has achieved a 99.54%
and TL-based MEC attained 98.76% scheduling accuracy. The result showed that
the proposed BMEC method is much better than the existing system. Therefore, the
proposed system has obtained both scheduling accuracy and security by reducing
computational delay. This shows that system is more scalable than conventional
systems.

6.6 CONCLUSION
In this work, the BMEC system is proposed and designed using a TL-based CBI
method. The TL-based CBI method is used to identify the clone block attack that was
attained in the BMEC system. This system consists of IoT devices, an Edge server, a
Cloud Server, and a Blockchain. The implementation of blockchain achieved a higher
security with a TL-based CBI method. This method predicted the various clone block
Efficient Blockchain-Based Edge Computing System 107

attacks in the edge server with a knowledge of legal block ID and Key functions.
The performance result showed that the total delay in BMEC gets increased non-
linearly, which achieved a high throughput transaction and low latency. Also, the
proposed BMEC system has obtained a scheduling accuracy of 99.54% compared to
the existing TL-based MEC system. This showed that this system is more scalable
than the conventional systems.

REFERENCES
[1] C. Xu, X. Yu, H. Cui, H. Qiu, Y. Huang, and Y. Xiao, “Privacy computing technology
and secure architecture based on blockchain,” Journal of Parallel and Distributed
Computing, vol. 145, pp. 68-80, 2020.
[2] J. Liu, T. Peng, and W. Wang, “Blockchain-based edge computing resource allo­
cation in IoT: A deep reinforcement learning approach,” IEEE Internet of Things
Journal, vol. 8, no. 6, pp. 4099-4108, 2021.
[3] W. Jin, Y. Xu, Y. Dai, and Y. Xu, “Blockchain-based continuous knowledge
transfer in decentralized edge computing architecture,” Electronics, vol. 12, no. 5,
p. 1154, 2023.
[4] K. Samy, Y. R. Rezk, M. A. R. Ghany, and A. A. R. Shehab, “A blockchain system
for task offloading security in MEC systems,” Journal of Network and Computer
Applications, vol. 158, p. 102660, 2020.
[5] Q. Tang, H. Zhu, Y. Luo, and J. Liu, “Secure aerial computing architecture using
blockchain,” IEEE Transactions on Network and Service Management, vol. 16, no. 3,
pp. 1047-1058, 2019.
[6] W. Wang, X. Li, X. Shen, and Y. Liu, “Triple real-time trajectory privacy protection
methodology in MEC,” IEEE Transactions on Mobile Computing, vol. 20, no. 1,
pp. 101-115, 2021.
[7] W. Wu, X. Yuan, Y. Zhang, and J. Zhang, “PBFT method for overcoming committee
selection issues and improving network security,” IEEE Transactions on Parallel and
Distributed Systems, vol. 32, no. 5, pp. 1034-1047, 2021.
[8] X. Wang, Y. Wang, Z. Wang, and Y. Sun, “Deep reinforcement learning (DRL) based
method using mean field theory (MFT),” IEEE Transactions on Neural Networks and
Learning Systems, vol. 31, no. 8, pp. 2901-2913, 2020.
[9] S. Zhang, X. Wang, T. Guo, and Z. Huang, “Blockchain-based deep actor-critic
method for task offloading,” IEEE Access, vol. 9, pp. 158205-158218, 2021.
[10] J. Yang, X. Wei, Z. Wang, and H. Zhao, “Blockchain-based conditional privacy­
preserving authentication method for vehicular edge computing,” IEEE Internet of
Things Journal, vol. 8, no. 3, pp. 1831-1841, 2021.
[11] S. Ma, T. Zhang, Y. Xu, and W. Wu, “BFT model to improve performance in MEC,”
IEEE Transactions on Computers, vol. 70, no. 7, pp. 1085-1099, 2021.
[12] S. Huang, X. Liu, J. Chen, and Q. Wu, “Service provider location using a blockchain
network,” Sensors, vol. 21, no. 5, p. 1652, 2021.
[13] P. Kumar, S. Singh, R. Misra, and Y. Liu, “SAE-ABIGRU: A mechanism for intru­
sion detection in MEC,” IEEE Transactions on Information Forensics and Security,
vol. 16, pp. 3005-3016, 2021.
[14] J. Liu, Z. Zhu, X. Hu, and Y. Zhang, “Edge-based networks for increased security
and efficiency,” IEEE Transactions on Dependable and Secure Computing, vol. 18,
no. 4, pp. 1770-1781, 2021.
108 IoT and Machine Learning for Smart Applications

[15] X. Huang, Y. Liu, W. Wang, and T. Zhang, “Hierarchical federated learning network
enabled by a multi-layer blockchain,” IEEE Transactions on Network and Service
Management, vol. 18, no. 1, pp. 100-113, 2021.
[16] P. Liu, L. Zhang, J. Zhou, and S. Wang, “P2P consortium blockchain-based trading
system for resources,” IEEE Transactions on Emerging Topics in Computing, vol. 9,
no. 2, pp. 769-782, 2021.
[17] S. Fugkeaw, T. Boonkrong, P. Jantaporn, and A. Sae-Tang, “IoT aggregation and
data transmission method using digital signing and encryption,” Future Generation
Computer Systems, vol. 116, pp. 164-175, 2021.
[18] Y. Zhong, Q. Wang, M. Li, and X. Tang, “A hybrid system for double-layer mining
service offloading,” IEEE Transactions on Cloud Computing, vol. 8, no. 1, pp. 14­
27, 2020.
[19] J. Ren, Y. Zhang, X. Huang, and T. Wang, “Task-offloading strategy leveraging
blockchain technology and trust-awareness,” IEEE Transactions on Sustainable
Computing, vol. 6, no. 4, pp. 593-606, 2021.
[20] F. Wang, Z. Liang, T. Chen, and Y. Sun, “Blockchain authentication scheme for
messages,” Journal of Network and Computer Applications, vol. 173, p. 102881,
2021.
[21] X. Du, L. Chen, Y. Liu, and R. Wang, “BIECS: Blockchain and intelligent edge
cooperation for resource sharing,” IEEE Internet of Things Journal, vol. 8, no. 10,
pp. 8555-8567, 2021.
[22] Y. Tian, Z. He, L. Li, and Y. Xu, “BSSMeta: A secure searching approach for meta­
data in MEC using blockchain,” IEEE Transactions on Services Computing, vol. 14,
no. 4, pp. 1371-1384, 2021.
[23] A. Chavhan, S. Patil, Y. Fang, and R. Zheng, “Security scheme for vehicle platooning
integrating edge computing and blockchain,” IEEE Transactions on Intelligent
Transportation Systems, vol. 23, no. 3, pp. 2034-2046, 2021.
[24] L. Qi, X. Liu, Z. Wang, and Y. Wu, “Cooperative proof-of-work (PoW) strategy
named Relay-PoW,” IEEE Transactions on Information Theory, vol. 67, no. 4,
pp. 2323-2338, 2021.
[25] B. Jiang, Q. Zhao, Y. Liang, and M. Li, “Attribute-based encryption technique,
blockchain, and smart contracts for access control,” IEEE Access, vol. 9, pp. 184123­
184137, 2021.
[26] C. Lang, Y. Huang, Z. Zhang, and S. Wei, “Cooperative carbon monoxide detection
using blockchain technology,” IEEE Transactions on Vehicular Technology, vol. 70,
no. 8, pp. 7860-7870, 2021.
[27] J. Li, X. He, Y. Zhang, and M. Chen, “Blockchain-assisted access control for
healthcare systems in fog computing,” IEEE Transactions on Emerging Topics in
Computing, vol. 9, no. 3, pp. 1264-1275, 2021.
[28] Y. Li, H. Zhao, X. Xu, and Y. Wang, “Blockchain-based data integrity verification
scheme for cloud-edge computing,” Future Generation Computer Systems, vol. 117,
pp. 341-351, 2021.
[29] R. Li, W. Yu, H. Zhao, and Y. Sun, “Secure blockchain-based access control
framework for IIoT,” IEEE Internet of Things Journal, vol. 8, no. 4, pp. 2623­
2636, 2021.
[30] X. Ding, Y. Liu, L. Li, and J. Yang, “IoT devices assistance from edge servers for data
analysis and storage-related tasks,” IEEE Transactions on Cloud Computing, vol. 9,
no. 4, pp. 1323-1334, 2021.
Efficient Blockchain-Based Edge Computing System 109

[31] L. Sun, X. Zhou, T. Chen, and Y. Gao, “Truthful auction for fog system (TAFS) to
incentivize fog nodes,” IEEE Transactions on Network and Service Management,
vol. 18, no. 2, pp. 1884-1896, 2021.
[32] X. Du, Y. Wang, Q. Zhang, and L. Chen, “Matching mechanism based on smart
contracts for renting associations,” IEEE Transactions on Smart Grid, vol. 12, no. 1,
pp. 567-578, 2021.
[33] H. Nguyen, X. Wang, T. Zhang, and P. Li, “Proof-of-reputation consensus mech­
anism based on block verification,” IEEE Transactions on Network and Service
Management, vol. 18, no. 1, pp. 104-116, 2021.
 Introducing a Compact
7 and High-Speed Machine
Learning Accelerator
for IoT-Enabled Health
Monitoring Systems
Kavitha V.P., Magesh V., Theivanathan G.,
and S.S. Saravana Kumar

7.1 INTRODUCTION
Recently, high-speed ML accelerators have evolved into new innovations due to
the arrival of IoT-enabled AI integration [1]. Recent trends in medical applications
have seen a paradigm shift toward edge and embedded devices, emphasizing the
need for high-performance hardware to meet the growing demands of real-time data
processing. As the Internet of Things (IoT) continues to expand its footprint across
industries, the integration of artificial intelligence (AI) at the edge has become a crit­
ical driver of progress [2].
High-speed machine learning accelerators are used in different fields like image
and signal processing. The hardware components in accelerators support for the inte­
gration of AI and ML programs [3]. Also, it allows an IoT environment for continuous
monitoring. The data is updated to the cloud continuously with the consumption
of huge amount of power. The collected data from the cloud is used for different
forecasting, such as weather, disaster, traffic, etc.
Recent trends in the tech industry highlight the increasing prevalence of IoT
ecosystems across domains such as health care, smart cities, industrial automation,
and autonomous vehicles. These trends make the need for efficient and high-speed
ML acceleration at the edge. The processors needed to work heavily when the volume
of the data increases [4]. As data volumes grow exponentially, the ability to process
and interpret this data directly on the device offers substantial advantages, including
reduced latency, enhanced privacy, and more robust operation, particularly in envir­
onments with limited connectivity.
In this work, design optimization of a High-Speed Machine Learning Accelerator
is proposed, which brings AI capabilities closer to the edge in IoT deployments. By
using cutting-edge techniques such as Support Vector Machines (SVM) for data

110 DOI: 10.1201/9781032623276-7

Introducing a Compact and High-Speed Machine Learning Accelerator 111

truncation and Honey Batcher optimization for hyperparameter tuning, our hardware
accelerator promises to be a game-changer in the IoT landscape. The performance
of the processors can be improved in terms of accuracy and efficiency for various
applications.
This work also explains optimization methodologies to illustrate how these
advancements contribute to the realization of intelligent and responsive IoT devices.
By applying the high-speed ML revolution, we prove that our work will give signifi­
cant advancements in IoT-enabled AI integration, opening up new possibilities for
innovation and problem-solving across multiple sectors.

7.2 RELATED WORK

Ali et al. [5] analyzed the performance of accelerators with different ML algorithms.
By varying the input size, the power consumption of the processors was analyzed
using various algorithms. The processor with a Field Programmable Gate Array
(FPGA) board shows superior performance in terms of delay and area.
The concept of prediction is used in many fields for future planning. The prediction­
based processor is designed by Braatz et al. [6]. The delay of a particular algorithm
is detected before the execution to save the power. The resources are allocated earlier
to reduce the overall delay of the circuits. Lee and Han [7] constructed a matrix
optimization-based hardware accelerator for low-power applications. The busy and
idle states of the program are calculated, and the functions of the matrix are optimized
with array segmentations.
Gentile and Serio [8] explored the concept of a new feature extraction algorithm
for the design of a big data processing hardware accelerator. The data from a huge
database is processed using a new feature extraction algorithm, and relevant data
is identified to avoid unnecessary processing of data. In Kee et al. [9], the author
designed a SVM-based hardware accelerator for environment processing. The sensor
data collected from the environment is processed by ML algorithms, and redundant
data is removed to avoid redundant storing of data. Experimental results show that the
power savings achieved up to 5% when applying ML.
Chen and Hao [10] proposed a neural network-based data fusion algorithm for
hardware accelerators. The neurons are processed to locate redundant data in the
processor, and a masking strategy is applied to eliminate unnecessary data. Song
et al. [11] developed an image-processing hardware accelerator for high-speed
applications. Then, the FPGA processor is used for image processing and analyzed
in terms of slice, look tables, and delay. Finally, the image segmentation algorithm is
loaded for power evaluations.
The concept of approximate computing is used to achieve low power in processors.
The approximate messaging approach was applied in communication systems by
Brennsteiner et al. [12] to achieve a low power. The data is handled effectively with
minimum error rates than other algorithms. Opportunistic redundancy is the approach
used in communication systems to improve data transmission reliability over unreli­
able or error-prone channels. The opportunistic redundancy-based hardware acceler­
ator is designed by Dong et al. [13] for signal processing. The overall area and time
requirements are optimized by the proposed approach.
112 IoT and Machine Learning for Smart Applications

Lai et al. [14] proposed a new attack detection mechanism for hardware
accelerators. The ML-trained mechanism is utilized to learn the attacker’s behavior
when compared to the normal users. The data from the Kaggle website is used to train
the ML model for attacker detection.
Zheng et al. [15] applied encryption algorithms in ML-based hardware accelerators.
The security of the processors can be improved by applying encryption algorithms in
accelerators. The homomorphic encryption is processed to secure data in accelerators
to prevent attackers’-learning is a reinforcement learning algorithm used in ML. It’s
designed to solve problems with minimum computational requirements. The concept
of q learning used in hardware accelerators by Sutisna et al. [16] for multimedia pro­
cessing applications.
Wang et al. [17] constructed a hardware accelerator using the strategy of paral­
lelism. The parallelism is used to minimize the execution time by applying simul­
taneous execution. The data is processed parallelly to avoid latency in execution.
Aboye et al. [18] proposed a new processor design using triple modular redundancy.
The fault-tolerant capability of the system can be improved by placing redundant
modules. These redundant modules help when an error is detected.
The reconfiguration of the system is used to switch over the functionality based on
user requirements. The reconfiguration-based accelerator is proposed by Vranjkovic
and Struharik [19] for multimedia applications. The power consumption of the pro­
cessor can be optimized by reconfiguration. Edge-enabled ML refers to the prac­
tice of deploying machine learning models and algorithms on edge devices. This
approach was applied by Suresh and Renu Madhavi [20] in hardware accelerators for
communications systems.
The hybrid Polyvalent ML-based hardware accelerator is proposed by Zhou et al.
[21]. It combines the different ML algorithms for processing the data in accelerators.
Based on applications, the ML algorithm can be loaded from memory. Aizaz et al. [22]
designed a new type of multiplier architecture for ML-based hardware accelerates.
The power consumption of processors is greatly optimized by introducing new multi­
plier structures. Likewise, Nurvitadhi et al. [23] proposed a new vector multiplication
architecture for ML-based accelerator design. Xin et al. [24] developed a new multi­
layer architecture for image-processing ML hardware accelerators. The pixels in the
processors are optimized using the multi-layer network.

7.3 PRELIMINARY

7.3.1 Multipliers
High-speed machine learning accelerators have become essential part all fields
for real-time AI integration and processing at the edge. The multiplier is the key
component of accelerators, which performs data multiplications and addresses
calculations.
Multipliers are specialized hardware units designed to perform the crucial task of
rapidly and efficiently executing complex mathematical operations. It plays a major
role in ML algorithms for difficult computations. The demand for high-speed and
low-power multipliers is increases when ML-based applications grow in all fields.
Introducing a Compact and High-Speed Machine Learning Accelerator 113

For high-speed machine learning accelerators, multipliers are the key components
responsible for executing the numerous matrix multiplications and convolution
operations essential for tasks such as image processing, NLP, pattern recognition,
and analysis. The performance of multipliers directly impacts the overall speed and
efficiency of an ML-based accelerator. Different algorithms and architecture have
been introduced by the researchers to design a multiplier in minimal area and delay.
The integration of the inelegance algorithm in the multiplier leads to area and delay
minimization achievements.

7.3.2 HBO
The honey badger’s remarkable foraging behavior serves as the inspiration for the
Honey Badger Optimization (HBO) model, a nature-inspired algorithm that aims to
solve complex optimization problems. This mathematical model is based on the two
key phases observed in the honey badger’s foraging process: the digging and honey
stages.

Initialization (Step 1): The HBO model begins by initializing a population of

candidate solutions. The population is represented as an array, where each element
corresponds to a potential solution. Each solution is represented as a vector, and
the population initialization involves generating these vectors within specified
bounds.

Candidate Solution Population (pop):

population(pop) = [X11 X12...XX2^ ...Xn1 Xn2... X1pX2p Xnp ] (7.1)

Honey badger Ithposition, x1 = [ xll, xp,..., xp ] (7.2)

In above equation, “pop” represents the entire population, and each “x_l” in Equation
(7.3) is an individual candidate solution as follows:

x, = lbl + RN1 x(ubl - lbl) (7.3)

where RN is the random number with upper and lower bounds.

Intensity Calculation (Step 2): In the HBO model, an intensity value (IV) is
calculated based on Inverse Square Law. This value represents the energy or distance
between the badger to prey.

Square Law expressed in Equation (7.4).

S
IV, = RN x----- - (7.4)
l 2 4nd,2
114 IoT and Machine Learning for Smart Applications

where
S=(xl-xl+1 ), that indicated the concentration strength
dl=xprey-xl, that specified a distance among prey and lth honey badger

Here, “S” represents the concentration strength, and “d_l” is the distance from source
to destination.

Density Factor Update (Step 3): a-density factor controls the time-varying ran­
domization, facilitating balance between convergence rate during the optimization
approach. This factor decreases over time, smoothing the solution.

t
C x expexp (7.5)
tmax
where
[max ^ allowable computation
C ^ Constant > 1 (default = 2)

In above Equation, “t_max” is the Max_No. of iterations, and “C” is a constant typ­
ically set to a value greater than or equal to 1.

Escaping Local Optima (Step 4): The HBO model introduces a mechanism to
escape local optima. A flag (F) is used to change the position, enabling a more exten­
sive exploration of the search space.

Position Update (Step 5): The location of badger is updated using the digging stage
and the honey stages.

= X + F x B x I x X + F x r3x^x d,l x
Xnewpreyprey
cos (2 n.rand4 )x [1 - coscos (2 n.rand5 )J (7.6)

where
Xpeey ^ Prey position
B ^ ability factor of Honey Badger

There is a flag (F) work is used to modify the direction of search that is expressed in
below Equation (7.12):

F = {1 if rand6 < 0.5 -1 else, (7.7)

In the phase of Honey stage, the beehive is attained by a honeyguide bird using honey
badger which is expressed as follows (7.8):

= X + F.rand7. ^ .d,l
Xnewprey (7.8)
Introducing a Compact and High-Speed Machine Learning Accelerator 115

In above Equation, “X_new” represents the new position, the position of the prey
denoted as “X_prey”, “F ’ is flag controlling the search direction, “p ’ represents ability
of honey_badger’s, “I” the intensity, “d_l” is the distance between best solutions, and
various random values are used for perturbation.
The HBO model encompasses these phases, combining exploration and exploit­
ation to find optimal solutions in the search space. By adapting the behavior of the
honey badger in foraging for solutions, the HBO model provides a unique and effi­
cient approach to solving complex optimization problems.

7.3.3 SVM
SVM are a powerful ML approach especially used for classification problems.
SVMs are mainly used where data points belong to two distinct classes and need to
be separated by a hyper_plane. The basic strategy of SVM, is to discover the most
advantageous hyper_plane that optimizes the distinction between the two classes,
leading to improved generalization and robustness.
The objective of SVM is to find the decision boundary that separates the data into
two classes. This decision boundary is represented by a hyperplane, which can be
expressed as:

Hyper_plane Equation: w * x + b = 0 (7.9)

where, “w’-weights or coefficients of the hyper_plane, “x’-feature vector of the data

point, “b’-bias term, and “f(x)’ decision function.
The decision function can be defined as follows:

Decision Function: f(x) = sign(w * x + b) (7.10)

In above Equation, “f(x)’ is the decision function, “sign’ is the sign function (indicates
the fall of data point on the hyper_plane), and “b’ is the bias term.

7.4 PROPOSED SYSTEM

This work proposes a new ML-based multiplier to improve the computational effi­
ciency of health care data processing while maintaining accuracy. The proposed
architecture also supports medical image processing. The proposed architecture
is best suited for applications where power consumption and resource utilization
are critical concerns. It combines two key elements: multiplier truncation and
compensation values and the selection of optimal values using SVM, as shown in
Figure 7.1.

Multiplier Truncation and Compensation

This aspect of the system involves the controlled reduction of the precision of
multipliers. The truncation of less significant bits of the multiplier leads to lower
computations and reduced power computations. However, this truncation introduces
errors in the results. The compensation circuits add compensation values to reduce
116 IoT and Machine Learning for Smart Applications

FIGURE 7.1 Proposed architecture for disease prediction.

that error. Compensation values are introduced to address the accuracy loss caused by
multiplier truncation. These values are determined through optimization techniques
with the goal of minimizing the impact of truncation errors. The choice of these com­
pensation values is critical to balancing efficiency and precision.
Here, the SVM is used to find out the best compensation values in order to reduce
an error value of multiplier during truncation. The error values of the multiplier
changed based on the number of ones present in the least significant bits of the multi­
plier output.
The system utilizes approximate multipliers in its architecture. Approximate
multipliers are designed to provide fast but slightly less precise results. These
multipliers are well-suited for applications where a small degree of error is accept­
able in exchange for significant improvements in computational speed and efficiency.
The design of a truncated multiplier includes three different stages: partial product
generation, truncation of bits, and final vector merging addition stages. In trunca­
tion, the hyperparameter-tuned SVM is applied for truncation and adds compensation
values to the multiplier final outputs.
By using approximate multipliers in hardware accelerators, the system achieves
a substantial improvement in computational efficiency. The proposed architecture is
best suited for ML-based hardware accelerators, especially in real-time signal pro­
cessing, edge computing, and IoT devices.
The reduced precision of multipliers and optimized compensation values lead to
savings in terms of hardware resources and power consumption. This is particularly
valuable in resource-constrained environments. While focusing on efficiency, the
system ensures that a satisfactory level of accuracy is maintained through the careful
selection of compensation values. This is achieved with the assistance of the SVM,
which helps strike a balance between computational speed and precision.
Introducing a Compact and High-Speed Machine Learning Accelerator 117

Hyperparameters of SVM
SVMs offer several hyperparameters that allow for fine-tuning the model to achieve
the best performance for a specific dataset. The primary hyperparameters include:
Kernel Function (K): SVMs can use different kernel functions to formulate data
into a large-dimensional space. The SVM kernel functions are classified as linear,
radial basis function (RBF), polynomial and sigmoid.
Regularization Parameter (C): The regularization parameter “C” reduces the clas­
sification error. The minimum value of “C” denotes a wider margin but may lead to
errors, while a larger “C” emphasizes accurate classification, potentially at the cost
of a narrower margin.
Kernel-specific Hyperparameters: If a kernel function is used (e.g., the RBF
kernel), there are additional hyperparameters specific to the kernel. For the RBF
kernel, these include “gamma,” which controls the shape of the decision boundary.

HBO-Based Hyperparameter Tuning

The HBO algorithm is a nature-inspired optimization technique that can be applied to
SVM hyperparameter tuning. HBO leverages the behavior of honey badgers in for­
aging for prey to explore and exploit the search space for optimal solutions.
In the context of SVM hyperparameter tuning, HBO can be used to find the best
combination of hyperparameters, such as the choice of kernel, “C,” and kernel­
specific hyperparameters like “gamma.” HBO mimics the honey badger’s ability to
balance exploration and exploitation to find the best solutions.
The need for HBO-based parameter tuning arises from the fact that
hyperparameters significantly impact SVM performance. Finding the right combin­
ation of hyperparameters can make the difference between a well-performing and a
poorly performing model. HBO’s ability to efficiently explore the hyperparameter
space can lead to improved model accuracy and robustness. The pseudocode for the
proposed multiplier construction is given below:

# Define the No. of honey badgers

population size = N
# Initialize a population of honey badgers with
random hyperparameters
population = initialize population(N)
# Define the objective function to be optimized
(e.g., classification accuracy)
def objective function(hyperparameters):
# Train an SVM model with the given
hyperparameters
model = train svm model(hyperparameters)
# Evaluate the defined model’s performance
accuracy = evaluate model(model, validation data)
return accuracy
# Set the maximum number of iterations for the HBO
algorithm
118 IoT and Machine Learning for Smart Applications

max iterations = max iter

# Set the initial exploration and exploitation
parameters
exploration param = initial exploration param
exploitation param = initial exploitation param
# Initialize the best fitness and corresponding
hyperparameters
best fitness = 0 # Initialize with a low value
best hyperparameters = None
# Main optimization loop
for iteration in range(max iterations):
# Calculate the fitness of each honey badger in the
population
for i in range(population size):
fitness = objective function(population[i] .
hyperparameters)
population[i].fitness = fitness
# Update the best fitness and
corresponding hyperparameters
if fitness > best fitness:
best fitness = fitness
best hyperparameters = population[i].
hyperparameters
# For each honey badger in the population
for i in range(population size):
# Update exploration and exploitation
parameters based on the iteration number
exploration param = update exploration
param(iteration, max iterations)
exploitation param = update exploitation
param(iteration, max iterations)
# Determine the direction of movement
(exploration or exploitation) for the honey badger
if random() < exploration param:
# Explore: Generate a random
hyperparameter within valid bounds
new hyperparameters = generate random
hyperparameters()
else:
# Exploit: Modify the current
hyperparameters
new hyperparameters = exploit
hyperparameters(population[i].hyperparameters,
exploitation param)
Introducing a Compact and High-Speed Machine Learning Accelerator 119

The process begins by defining the population size, which determines the number
of “honey badgers” in the population. These honey badgers represent sets of
SVM hyperparameters. The population is initialized with random hyperparameter
configurations, providing a diverse starting point for optimization. The objective
function plays a central role. It measures the performance of the SVM model with
a given set of hyperparameters. Typically, this could be a classification accuracy
metric on a validation dataset. The objective function is essential because the HBO
algorithm’s objective is to maximize it. The better the model performs, the higher the
fitness value assigned to the hyperparameters.
The heart of the algorithm lies in this main optimization loop. The loop runs for
a predefined number of iterations (max_iterations). In each iteration, the algorithm
evaluates the fitness of each honey badger in the population. Fitness is determined
by applying the objective function to each set of hyperparameters, reflecting how
well the SVM model performs with those hyperparameters. The algorithm maintains
awareness of the best-performing set of hyperparameters. As fitness values are
computed, the algorithm compares each honey badger’s fitness to the best fitness
observed. If a honey badger’s hyperparameters yield better results, the best fitness is
updated, and the corresponding set of hyperparameters is recorded as the best solu­
tion found.
The HBO algorithm distinguishes between exploration and exploitation phases
controlled by the coefficient factors. These parameters are altered during every iter­
ation in order to balance the exploration and exploitation ranges. In each iteration the
algorithm decides whether to explore or exploit each honey badger. If exploration is
favored, the algorithm generates random hyperparameters within valid bounds.
In contrast, during exploitation, it modifies the current hyperparameters. This
choice is guided by the exploration and exploitation parameters. The generated
or modified hyperparameters are checked to ensure they fall within valid bounds.
For instance, in the case of SVM hyperparameters, the regularization parameter
(C) should be positive, and the kernel-specific parameters should also adhere to
appropriate constraints. If the newly generated or modified hyperparameters improve
the fitness of a honey badger, the hyperparameters are updated for that honey badger.
This process mimics the honey badger’s ability to adapt and refine its approach as
it searches for prey. After the optimization process concludes, the algorithm returns
the set of hyperparameters associated with the honey badger that achieved the best
fitness. These hyperparameters represent the optimal configuration for the SVM
model, which should lead to the best performance on your specific problem.

7.5 EXPERIMENTAL RESULTS

The results of our proposed system are implemented in Xilinx 12.1 using Verilog.
This programming is used to compare the resources between existing and proposed
systems. The proposed model is compared in terms of the slice, logic tables, and
delay. When comparing the slices, the proposed model shows a smaller slice than the
other model. The proposed system demonstrated a marked reduction in logic elem­
ents employed and shows the more reasonable utilization of area in the FPGA board.
The “LUT” metric is used to evaluate the logical utilization of the proposed model
120 IoT and Machine Learning for Smart Applications

FIGURE 7.2 Product output.

FIGURE 7.3 Final output.

with other models. These findings underscore the positive impact of our approach,
which leverages SVM integration in FPGA design. The simulated waveforms are
given in Figures 7.2 and 7.3.
The observed resource efficiency of SVM is very high when compared to existing
systems. These results prove that more streamlined utilization of logic elements and
combinatorial logic resources. The reduction in resource consumption can be attributed
to the optimized precision of multipliers and careful compensation strategies. This
outcome underscores the suitability of our approach for resource-constrained envir­
onments where the efficient allocation of hardware resources is paramount.
Moreover, our system’s resource efficiency not only conserves FPGA resources
but also holds the potential for enhanced power efficiency. As FPGA devices increas­
ingly find applications in edge computing and IoT contexts, the system’s ability to
balance computational speed and precision holds particular relevance.
The performance of the proposed system is given in Table 7.1. Slice is a metric
in FPGA terminology that refers to a group of configurable logic elements. A lower
number of slices in the “With SVM” configuration compared to “Conventional” and
“Without SVM” suggests that the implementation of SVM is more resource-efficient in
Introducing a Compact and High-Speed Machine Learning Accelerator 121
122 loT and Machine Learning for Smart Applications
Introducing a Compact and High-Speed Machine Learning Accelerator 123

TABLE 7.1
Performance Analysis

S. No Parameter Conventional Without SVM With SVM

1 Slice 1219 612 505

2 LUT 2095 1106 856

Slice analysis
1400

FIGURE 7.6 Area analysis.

FIGURE 7.7 Delay analysis.

124 IoT and Machine Learning for Smart Applications

TABLE 7.2
Power Analysis

S. No. Parameter Conventional Without SVM With SVM

1 Power (nW) 12 9.5 7.6

Power

FIGURE 7.8 Power analysis.

terms of logic elements. Look-up tables (LUTs) are used for implementing combina­
torial logic. The number of LUTs used in the “With SVM” configuration is lower than
in both the “Conventional” and “Without SVM” configurations. This indicates that the
SVM-based system is more resource-efficient in terms of logic resources, as shown in
Figures 7.4 and 7.5. The comparison is graphically shown in Figures 7.6 and 7.7.
Table 7.2 shows the power analysis of the proposed model and it’s clear that the
proposed model “With SVM” consumes less power (7.6 nW) than the model “Without
SVM” (9.5 nW). The power analysis is graphically shown in Figure 7.8.

7.6 CONCLUSION
In conclusion, the HBO algorithm presents a nature-inspired approach to the challen­
ging task of tuning SVM hyperparameters. This optimization method leverages the
behaviors of honey badgers in foraging for prey, balancing exploration, and exploit­
ation to find the best set of hyperparameters for an SVM model.

REFERENCES
[1] M. Capra, B. Bussolino, A. Marchisio, M. Shafique, G. Masera, and M. Martina, “An
updated survey of efficient hardware architectures for accelerating deep convolutional
neural networks,” Future Internet, vol. 12, no. 7, p. 113, Jul. 2020.
Introducing a Compact and High-Speed Machine Learning Accelerator 125

[2] F. Cardells-Tormo, P.-L. Molinet, J. Sempere-Agullo, L. Baldez, and M. Bautista-

Palacios, “Area-efficient 2D shift-variant convolvers for FPGA-based digital image
processing,” in Proc. Int. Conf. Field Program. Log. Appl., Tampere, Finland, 2005,
pp. 578-581.
[3] L. Cavigelli, D. Gschwend, C. Mayer, S. Willi, B. Muheim, and L. Benini, “Origami: A
convolutional network accelerator,” in Proc. 25th, Ed., Great Lakes Symp. (VLSI),
New York, NY, USA, May 2015, pp. 199-204.
[4] S. Chakradhar, M. Sankaradas, V. Jakkula, and S. Cadambi, “A dynamically config­
urable coprocessor for convolutional neural networks,” in Proc. 37th Annu. Int. Symp.
Comput. Archit. (ISCA), New York, NY, USA, 2010, pp. 247-257.
[5] D. Ali, A. U. Rehman, and F. H. Khan, “Hardware accelerators and accelerators for
machine learning,” 2022 International Conference on IT and Industrial Technologies
(ICIT), Chiniot, Pakistan, 2022, pp. 01-07, DOI: 10.1109/ICIT56493.2022.9989124.
[6] Y. Braatz, D. S. Rieber, T. Soliman, and O. Bringmann, “Simulation-driven latency
estimations for multi-core machine learning accelerators,” 2023 IEEE 5th International
Conference on Artificial Intelligence Circuits and Systems (AICAS), Hangzhou, China,
2023, pp. 1-5, DOI: 10.1109/AICAS57966.2023.10168589.
[7] S. C. Lee and T. H. Han, “A 4-way matrix multiply unit for high throughput machine
learning accelerator,” 2019 International SoC Design Conference (ISOCC), Jeju,
Korea (South), 2019, pp. 113-114, DOI: 10.1109/ISOCC47750.2019.9078493.
[8] U. Gentile and L. Serio, “A machine-learning based methodology for performance
analysis in particles accelerator facilities,” 2017 European Conference on Electrical
Engineering and Computer Science (EECS), Bern, Switzerland, 2017, pp. 90-95,
DOI: 10.1109/EECS.2017.26.
[9] Minkwan Kee, Seung-jin Lee, Hyun-su Seon, Jongsung Lee, and Gi-Ho Park,
“Intelligence boosting engine (IBE): A hardware accelerator for processing sensor
fusion and machine learning algorithm for a sensor hub SoC,” 2017 IEEE Symposium
in Low-Power and High-Speed Chips (COOL CHIPS), Yokohama, Japan, 2017,
pp. 1-3, DOI: 10.1109/CoolChips.2017.7946377.
[10] H. Chen and C. Hao, “Hardware/software co-design for machine learning accelerators,”
2023 IEEE 31st Annual International Symposium on Field-Programmable Custom
Computing Machines (FCCM), Marina Del Rey, CA, USA, 2023, pp. 233-235,
DOI: 10.1109/FCCM57271.2023.00058.
[11] J. B. Song, Y. Kim, M. Lee, S.-S. Lee, and K. Kim, “A flexible machine vision ai
system for edge-oriented deep learning accelerators,” 2023 International Technical
Conference on Circuits/Systems, Computers, and Communications (ITC-CSCC), Jeju,
Republic of Korea, 2023, pp. 1-4, DOI: 10.1109/ITC-CSCC58803.2023.10212811.
[12] S. Brennsteiner, T. Arslan, J. S. Thompson, and A. McCormick, “A machine
learning enhanced approximate message passing massive MIMO accelerator,”
2022 IEEE 4th International Conference on Artificial Intelligence Circuits and
Systems (AICAS), Incheon, Republic of Korea, 2022, pp. 443-446, DOI: 10.1109/
AICAS54282.2022.9869942.
[13] B. Dong, Z. Wang, W. Chen, C. Chen, Y. Yang, and Z. Yu, “OR-ML: enhancing reli­
ability for machine learning accelerator with opportunistic redundancy,” 2021 Design,
Automation & Test in Europe Conference & Exhibition (DATE), Grenoble, France,
2021, pp. 739-742, DOI: 10.23919/DATE51398.2021.9474016.
[14] Y.-K. Lai, K.-P. Chang, X.-W. Ku, and H.-L. Hua, “A machine learning accelerator
for DDoS attack detection and classification on FPGA,” 2022 19th International SoC
Design Conference (ISOCC), Gangneung-si, Republic of Korea, 2022, pp. 181-182,
DOI: 10.1109/ISOCC56007.2022.10031506.
126 IoT and Machine Learning for Smart Applications

[15] M. Zheng, F. Chen, L. Jiang, and Q. Lou, “PriML: An electro-optical accelerator for
private machine learning on encrypted data,” 2023 24th International Symposium
on Quality Electronic Design (ISQED), San Francisco, CA, USA, 2023, pp. 1-7,
DOI: 10.1109/ISQED57927.2023.10129302.
[16] N. Sutisna, Z. N. Arifuzzaki, I. Syafalni, R. Mulyawan, and T. Adiono, “Architecture
design of Q-learning accelerator for intelligent traffic control system,” 2022
International Symposium on Electronics and Smart Devices (ISESD), Bandung,
Indonesia, 2022, pp. 1-6, DOI: 10.1109/ISESD56103.2022.9980698.
[17] C. Wang, L. Gong, X. Li, and X. Zhou, “A ubiquitous machine learning acceler­
ator with automatic parallelization on FPGA,” IEEE Transactions on Parallel and
Distributed Systems, vol. 31, no. 10, pp. 2346-2359, 1 Oct. 2020, DOI: 10.1109/
TPDS.2020.2990924.
[18] D. Aboye et al., “PyRTLMatrix: An object-oriented hardware design pattern for proto­
typing ML accelerators,” 2019 2nd Workshop on Energy Efficient Machine Learning
and Cognitive Computing for Embedded Applications (EMC2), Washington, DC,
USA, 2019, pp. 36-40, DOI: 10.1109/EMC249363.2019.00015.
[19] V. Vranjkovic and R. Struharik, “Coarse-grained reconfigurable hardware accelerator
of machine learning classifiers,” 2016 International Conference on Systems, Signals
and Image Processing (IWSSIP), Bratislava, Slovakia, 2016, pp. 1-5, DOI: 10.1109/
IWSSIP.2016.7502737.
[20] Suresh, B. N. Reddy and C. Renu Madhavi, “Hardware accelerators for edge enabled
machine learning,” 2020 IEEE REGION 10 Conference (TENCON), Osaka, Japan,
2020, pp. 409-413, DOI: 10.1109/TENCON50793.2020. 9293918.
[21] S. Zhou et al., “ParaML: A polyvalent multicore accelerator for machine learning,”
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems,
vol. 39, no. 9, pp. 1764-1777, Sept. 2020, DOI: 10.1109/TCAD.2019.2927523.
[22] Z. Aizaz, K. Khare, and A. Tirmizi, “FASBM: FPGA-specific approximate sum-based
booth multipliers for energy efficient hardware acceleration of image processing and
machine learning applications,” 2023 IEEE 31st Annual International Symposium on
Field-Programmable Custom Computing Machines (FCCM), Marina Del Rey, CA,
USA, 2023, pp. 210-210, DOI: 10.1109/FCCM57271.2023.00038.
[23] E. Nurvitadhi, A. Mishra, and D. Marr, “A sparse matrix vector multiply accelerator for
support vector machine,” 2015 International Conference on Compilers, Architecture
and Synthesis for Embedded Systems (CASES), Amsterdam, Netherlands, 2015, pp.
109-116, DOI: 10.1109/CASES.2015.7324551.
[24] G. Xin, Y. Zhao, and J. Han, “A multi-layer parallel hardware architecture for homo­
morphic computation in machine learning,” 2021 IEEE International Symposium
on Circuits and Systems (ISCAS), Daegu, Korea, 2021, pp. 1-5, DOI: 10.1109/
ISCAS51556.2021.9401623.
 Realization of Smart City
8 Based on IoT and AI
Pavithra N., Sapna R., Preethi,
Manasa C.M., and Raghavendra M. Devadas

8.1 INTRODUCTION
The term “smart city” is quite popular since it improves urban residents’ quality of life
by integrating many fields, like smart communities, smart transit, smart healthcare,
smart parking, and many more. Real-time data processing and the ability to generate
smarter decisions pose a substantial threat to the continuous evolution of complex
metropolitan networks. The bounds of conventional networks have been widened by
the phenomenal expansion of the many devices linked to the network. This significant
development introduced the third wave of the internet, followed by social networking
and static pages web (WWW), “Internet of Things (IoT)” [1]. Internet of Technology,
prominent field which can identify as well as share data across heterogeneous objects
that may be uniquely addressed. IoT has attracted interest from a variety of groups
because of the expansion of embedded systems and the hasty growth in the number
of devices being used. The Internet of Things (IoT) concept has expanded due to the
attention of multiple interest groups and advancements in embedded device tech­
nology. This develops useful efforts, including smart housing, smart cities, smart
health, etc. To effectively fulfill service requests, a huge volume of instantaneous data
processing is required. General data processing and analytical procedures are unable
to meet the demand for processing data instantaneously because of the enormous
expansion in data volume. Therefore, it is thought that working with analysis is the
best way to start building a smarter city. Academic and industrial professionals have
made numerous attempts to realize the concept of intelligent cities. The present-day
state of the art [2] documents numerous initiatives on particular areas of concern,
such as water resource management, waste management, parking management, and
many more. Big Data analytics is integrated with metropolitan IoT to realize the
smart city vision. As an illustration, a smart meter placed in a home gathers readings
from the meter and compares them to a predetermined power consumption limit.
Based on the comparison, the smart grid is notified of the energy demand that exists
at that moment. Consumers are simultaneously informed of the level of energy usage,
enabling them to effectively manage energy use. Although the buzzword “smart city”
has attained popularity in the era of technology, actual implementation is still in its
early stages. Multiple initiatives are done in this regard to build a practical smart city.

DOI: 10.1201/9781032623276-8 127

128 IoT and Machine Learning for Smart Applications

The heterogeneous data will be stored and processed using Hadoop [3]. After
Hadoop processing, intelligent decision-making linked to smart city functions
is produced. The decisions’ accompanying actions or events are then carried out.
Numerous state administrations are embracing the urban communities and executing
large information applications to arrive at the necessary degree of maintainability and
work on the expectations for everyday comforts. Urban areas use various advances
to work on the more elevated levels of solace to their residents. Shrewd urban areas
include decreasing asset utilization and costs notwithstanding more effectively and
successfully captivating their residents. One of the new advancements upgrades a huge
amount of information in urban areas. The gathering of enormous volumes of data
that can be used in useful fields has become a part of daily life. Success in numerous
industries and sectors relies heavily on the utilization and effective analytics of huge
quantity of data, commonly referred to as big data. Ninety percent of the world’s
digital data has been acquired in the previous 5 years alone, and big data growth is
expected to exceed 40 percent every year [4]. As a result, numerous governments
have started utilizing big data to support the growth and sustainability of smart cities.
As a result, cities were able to maintain the standards, requirements, applications, and
concepts of smart cities. Big data stores a lot of information, whether it is organized
or not and whether it is structured. Big data is a typical database that comes through
data processing. Large volumes of data will be efficiently stored by big data systems,
which will then supply information to enhance the services that smart cities deliver.
Using this information, big data will assist administrators in planning for any expan­
sion in smart cities, locations, and resources. Many cities are transforming into smart
cities because of advantages such as those for the environment, economy, and ana­
lysis. Therefore, we’ll talk about some of the advantages and possibilities that could
aid in transforming our city into a smart city. Through these advantages and possibil­
ities, we may redesign our city to be a smart city. We can reach higher standards of
sustainability, governance, and resilience by making use of benefits and possibilities.
We can enhance natural resource management and life quality by implementing intel­
ligent infrastructure management. Figure 8.1 illustrates the broader list of applications
based on smart cities using big data analytics.
The following are some advantages of having a smart city: Effective resource man­
agement: As resources become either limited or very expensive, they must be used
more carefully and integrated into solutions. Systems involving technology like cor­
porate resource planning with geospatial information systems will aid this process.
Locating waste distribution locations with a monitoring system will be easier, which
will also help with energy conservation and resource management. Applications for
interconnection and data gathering in smart cities that work together to provide ser­
vices and applications.

8.1.1 Greater Quality of Life

Smart cities will have enhanced lifestyles because of vast housing options, services,
and employment opportunities. The outcome is better and faster transit, adequate
knowledge to make decisions, and better living and working conditions at a location.
Realization of Smart City Based on IoT and AI 129

FIGURE 8.1 Overview of smart city application.

For these advantages to be realized, it is essential to establish high standards for

the quality, privacy, control, and security of the data. It is also vital to employ data
documentation standards to offer guidance on how to use content and data sets. This
involves a high level of sophistication in the application as well as the involvement of
people and resources.

8.1.2 Smart Education

A knowledge-based society will emerge, boosting capability and competitiveness.
Big data and information and communication technologies will play a part in this. Big
data in learning focuses on gathering information from students, teachers, parents,
administrators, infrastructure like schools, library systems, educational locations,
teachers, museums, and universities, as well as specialized data like tests, documents,
finance, and assessments. It also involves collecting information from institutions
such as universities. Data will develop into a valuable resource for information
extraction and analysis to improve education. Data, for instance, aids learning for
oneself when studying.

8.1.3 Intelligent Traffic Lights

Issues with traffic, pollution, and the economy all develop with population growth.
This is one of the most important strategies for managing heavy traffic and conges­
tion in smart cities. Traffic grids about traffic patterns are coupled to smart signals
and lights. Different aspects of traffic flow, such as traffic jams, vehicle speeds, and
waiting times at traffic lights, are detected by sensors. These parameters are provided
by the system, along with lights and signals.
130 IoT and Machine Learning for Smart Applications

8.1.4 Smart Grid

An essential element of smart cities. In order to gather data from electrical grid
networks. The reliability, affordability, and efficiency of electric power are all
improved. Smart remote controllers utilize two-way communication technology.
The distribution, production, and transmission systems must have smart sensors and
meters installed to get real-time data on current power usage and problems. It can
give customers almost real-time energy information and let them buy what they need
at reasonable pricing. Consumer electronics like washing machines and water heaters
can cost more to automatically control during a sale.

8.2 CUTTING-EDGE TECHNOLOGIES

It is now simpler than ever to communicate information across platforms due to
the IoT, which forms the foundation for sensors and devices to interact with each
other in a smart-cities context. To enhance the effectiveness of smart cities, we
may incorporate a smart city as a real system that connects many lives, including
transportation and buildings. Due to the growing usage of many communication
devices, IoT gadgets are positioned as the upcoming breakthrough technology by
harnessing all the capabilities afforded by Internet technology. The development
of city, like power grids, supermarkets, housing, water, transit, and buildings has
recently become smart with the use of the IoT [5]. As part of the smart-city program,
sensors and gadgets are integrated into hospitals, power grids, railways, pipelines,
roads, houses, infrastructure, bridges, and other things all over the world. Storage
facilities and strong computing capabilities are needed to support these types of
smart applications. One method to provide such a platform is to use the various
advantages of cloud services to enable data analytics in smart-cities management.
Figure 8.2 illustrates the overall system architecture of a smart city using analytics
of large volumes of data.
The massive amounts of data gathered from the current smart cities need to be
precisely evaluated to create effective and universal smart city architecture. In a smart
home or smart city ecosystem, sensors can be positioned in numerous areas to collect
data. The ecosystem of an urban city is designed and planned, which can benefit
from the offline processing of big data. However, it is useless for making decisions
in the present. Numerous methods are built on the Hadoop ecosystem, where smart
cities use big data systems to process, store, and mine data effectively. This helps
to improve many smart-city applications tremendously. Big Data can indeed assist
experts in designing and constructing smart-city services, assets, and infrastructure.
Big data has a broad range of qualities that highlight its tremendous potential for
growth and improvement.
The availability of cutting-edge technologies and techniques, however, places
limits on the seemingly limitless possibilities. Huge volumes of data can attain their
objective and improve processes in smart cities by utilizing proper methods and tools
for effective and accurate analysis of data. Such efficiency would facilitate stake­
holder cooperation and communication and make it easier to establish new facilities
and technology that can improve the smart city. Big data applications could grant
Realization of Smart City Based on IoT and AI 131

FIGURE 8.2 The system architecture.

many different areas of a smart city, enhancing customer experiences and services
while also helping businesses operate more effectively. Healthcare can be enhanced
by providing more patient attention, therapeutic and diagnostic instruments, manage­
ment of medical records, and preventive care services. Big data has the potential to
significantly increase environmental friendliness, adjust to fluctuating demand, and
optimize routes and schedules for transportation networks. Big data is precisely what
results from the production of numerous resources from databases. Smartphones,
GPS, environmental sensors, computers, and even people’s numerous apps, including
games, advertising applications, digital photographs, social media sites, and greater
numbers of faster data from previous times. Big data is a typical database that comes
through data processing.
Big data is characterized by variety, value, velocity, and veracity. Large amounts
of data will be efficiently stored by big data systems to supply knowledge that will
enhance the services provided by smart cities. Big data will use this information
to assist administrators in planning for any growth in smart locations, buildings,
and resources. To provide a framework and coding tools for the parallel computing
of huge datasets over numerous clusters, Hadoop was developed. MapReduce and
Hadoop Distributed File System, the two main parts of Hadoop, are interconnected
[6,7]. Despite the requirement for live/real-time data processing and storing in smart
cities, the streaming style will provide good connectivity among various sensors
inside the network. Such technology has lately become widely employed with the
launch of different real-time data processing platforms, like Apache-Sturm, Apache-
S4, Spark-streaming, and MLLib [7], which may permit data storage plus analysis
132 IoT and Machine Learning for Smart Applications

via many networked nodes/clusters. “Cloud computing” refers to a variety of alter­

native computing architectures with several nodes/clusters linked by a real-time
interaction system [8]. Large-scale, challenging, and time-consuming computer
tasks, such as mining vast amounts of data from social networks generated by
mobile applications, can be performed using cloud computing services [9]. IoT can
be connected with cloud computing services like infrastructure-as-a-service (IaasS),
software-as-a-service (SaaS), and platform-as-a-service (PaaS) [10]. This combin­
ation has the power to alter every business since big data technology makes it simple
to analyze large data. Cloud computing can also offer a virtualized computing ser­
vices platform that integrates security tools, data management tools, analysis soft­
ware, visualization platforms, and customer deployment [10]. With the help of the
cost-based paradigm that cloud computing may provide, users and organizations
will be able to access data whenever they need it, from any location. Through the use
of big data technology like the Hadoop framework, cloud computing also supplies
the underlying engine.

8.3 EMERGENT TECHNOLOGIES FOR COMMUNICATION

Emerging technology for communication can include both highly visible examples,
such as social media and smartphones, and less evident or hidden ones, such as speech
recognition software or drones. Even though they are evolving quickly, the majority
create a number of moral issues when made available to the general public. Emerging
technologies are, by definition, novel, cutting-edge, and still under development, but
they are anticipated to have significant socioeconomic effects. The user can typically
communicate information or news through communication technology; however, it
does not need to be immediate. That takes care of the essentials, but it’s still a some­
what nebulous description that doesn’t cover everything.
Smart cities are possible because of the smart networks that connect the user’s
gadgets, including cars, smart homes, and smartphones which are necessary for big
data analytics. Data analytics should be able to be used by the smart network to
receive and process data in an appropriate manner. Also, immediate responses should
be sent to many organizations pertaining to smart cities. As the data grows in real-time
applications, maintaining the quality of service (QoS) in smart networks becomes
an important feature. Recently, distributed application events in these applications
should be communicated in real time to the processing location. The events that
follow may be sent straight from their sources, unfiltered and unaggregated, or dir­
ectly as raw events [11].
IoT technology, which makes it possible for a wide range of objects to be detected
and remotely controlled over current network infrastructure, must be put into practice
if the smart city idea is to be adopted. This project will make opportunities for the
adaptable integration of various smart city components. The addition of sensors and
actuator components to the IoT will improve its accuracy and efficiency while also
producing financial rewards.
An IoT technology will be the best option to accomplish this, as was previously
stressed. Organizations like the Extensible Messaging Presence Protocol project,
through its foundation like XMPP Standards Foundation, build a platform for IoT
Realization of Smart City Based on IoT and AI 133

technology that is unconnected to any firm or cloud services. For instance, REST
technology provides a scalable architecture that facilitates device interaction via the
hypertext-transfer protocol (HTTP) and is tailored for IoT applications to allow com­
munication between any device and a server [11].

RFID
Radio Frequency Identification [3] uses electromagnetic waves to recognize and
track through tags affixed to things automatically. The tags, which come in active,
passive, and battery-assisted passive varieties, retain information electronically.
The battery-operated active kind periodically transfers its ID signal. The battery-
free passive versions employ the radio-electric energy sent by the reader. With a
small onboard battery, the passive battery-assisted model can only communicate
when an RFID reader is nearby. RFID is a widely used embedded communica­
tion technology that may be used to identify almost everything, including people,
clothing, pets, and even other people. Because of its wide range of applications,
RFID is now perfect for smart cities; it may be used for cargo tracking, hospitals,
and libraries.
The software manages and keeps track of the RFID tags that are added to user data.
It is either a piece of unique software or a smartphone app. RFID software frequently
includes a clever mobile app that functions in conjunction with it. This program can
connect via Bluetooth or beacon technology with the RFID reader. With another elec­
tronic RFID-enabled gadget, communication is possible. The item with which you
want to connect must have an RFID tag installed so that it can recognize the electro­
magnetic waves sent out by the RFID reader. The technology allows us to connect
two RFID-enabled devices, one of which is an RFID reader and the other of which
is an electronic gadget with an RFID tag. Some RFID readers are manufactured with
built-in RFID antennas, whereas others include antenna ports. To use the readers, the
users must add additional antennas [12].
Components of RFID used are as follows:

RFID tag: RFID tags come in passive and active varieties. A passive RFID tag is the
barcode at the store. It doesn’t require a power source, is easy to activate, and is paired
with a certain object. Like the sensor tag behind the library book, an active RFID tag
has a microchip that records data about the item and might also have an antenna or
onboard sensor.

RFID scanner: A device that reads RFID tags and gathers evidence concerning the
item or device to which the tag is attached is called an RFID scanner or reader. These
tethered or portable readers support both USB and Bluetooth. Although not every
barcode scanner has the capability to interpret an RFID tag, all RFID readers possess
the ability to read a barcode. We can classify the RFID readers into three groups based
on the range they cover, and they are described below.

• Low-Frequency (LF) RFID: When put within the range of 30 to 300 kHz,
RFID-enabled devices can communicate with low-frequency RFID readers.
They travel a lot less distance.
134 IoT and Machine Learning for Smart Applications

• High-Frequency (HF) RFID: High-Frequency RFID scanners are able to

communicate with RFID-capable gadgets that operate between 3 and 30 Mhz.
• Ultra-High-Frequency (UHF)-RFID: The strongest RFID readers are ultra-
high frequency (UHF) scanners because they can operate over a significantly
wider range (up to 300 MHz).

RFID Software
This manages and keeps track of the RFID tags that are attached to the user’s assets.
It is only run by a piece of proprietary software or a mobile app. RFID software fre­
quently includes a mobile app that functions in conjunction with it. This can com­
municate with the reader using Bluetooth or beacon technologies. Unlike barcodes,
which are limited to certain products, RFID tags are universally compatible and their
signal may be detected by a scanner even when they are covered.
Due to its potential, RFID is currently setting the standard for the creation of
integrated gadgets that are fully trackable. For example, a meter can become a smart
measuring device when an RFID tag is installed. RFID tags are used by many com­
panies. For instance, the development of an automobile’s assembly line while it is
being produced can be observed. Animals can be positively identified thanks to the
RFID microchips installed in livestock and companion animals. Through RFID-
embedded technology, such an identification approach makes intelligent equipment
for smart cities possible.

Wireless Sensor Networks (WSNs): It is a collection of wireless communication

technologies and distributed autonomous sensing nodes that use low-power integrated
circuits to disseminate data within the linked sensors. Various inexpensive, low-power,
small devices that are coupled with one or more sensors are supported by this net­
work. The sensor includes a radio transceiver for transmitting and receiving signals,
a microprocessor, an electronic circuit for interacting with the sensing devices, and
an energy source. The prospect of employing a sensor network with many intelligent
sensors has increased because of the WSN’s capacity to connect low-cost and small
devices [13]. This network makes it possible to make smooth connections and share
information between many contexts, which is efficient.
This feature of WSN makes it applicable to numerous fields, including water
quality monitoring, machine health monitoring, and industrial process monitoring
and control. The WSN is suitable for smart city integration since it can handle large-
scale deployment in any setting [14]. The network provides a quick, affordable way
to set up scattered monitoring and control equipment, avoiding the possible over­
head expenses of wired systems. Real-time monitoring of physical and environ­
mental factors including temperature, pressure, light, and humidity is possible with
the WSN. These ambient variables are controlled by devices, like switches, motors,
or actuators, using an effective wireless connection. These features make WSN meant
for smart buildings, smart homes, and smart health. Despite the difficulties that WSN
encounters regarding energy use.

Ultra-wideband, Zig Bee, and Bluetooth, Wireless Fidelity (WiFi) [16]: If a city
has integrated wireless communication platforms, it might be referred to as smart.
Realization of Smart City Based on IoT and AI 135

A rapidly expanding technology, wireless communication offers rising levels of

flexibility and mobility. Dynamic network development, low cost, and simple imple­
mentation are offered by wireless technology. Users can substitute traditional cable
networks and access the Internet at broadband speeds when they are linked to an
access point or using WiFi in ad hoc mode. Ultra-wideband technology focuses on
high-bandwidth indoor short-range wireless networks with multimedia connectivity.
ZigBee is designed for short-range wireless communication and has the option of a
long battery life [15].
The goal of the Bluetooth standard is to replace bulky, expensive cords with low­
cost, short-range devices for computer input-output devices like printers, mouse,
keyboards, and any others. These short-range wireless technologies, which require a
low-power network, have been essential for wireless data transmission. However, the
QoS offered by wireless communication technology ought to improve because spe­
cific electronic devices have a strong leaning to travel from one network to another
without encountering communication problems [16]. The creation of self-healing,
self-organizing wireless communication networks is crucial to the success of a smart
community.

5G, Long-Term Evolution Advanced (LTE-A), 4G Long-Term Evolution (LTE)

and 5G [17]: The 4G wireless network, a development of the current 3G wireless
standards, is called LTE technology. By extending this concept, 4G transforms hybrid
data and voice networks into data-only IP networks. Additionally, 4G combines
orthogonal frequency division multiplex (OFDM) and multiple input multiple output
(MIMO) to increase data throughput over 3G [17]. A transmission method called
OFDM makes use of several widely spaced carriers that are modulated at low data
rates. This method uses a spectral efficiency system that allows for large data rates
and the sharing of a single channel by several users. MIMO employs several antennas
at the transmitter and receiver to significantly increase data throughput and spec­
tral efficiency. It is anticipated that most of the machine-to-machine communication
traffic will use 4G wireless networks.
By introducing high bandwidths, LTE-Advanced (LTE-A) enables the transition
from fourth generation to fifth generation. It promises speeds that are nearly three
times as fast as those of the basic LTE network, as well as includes features like
carrier aggregation, more MIMO, coordinated multipoint, relay stations, and diverse
networks. Furthermore, A more advanced technology known as 5G supports a band­
width of up to 10 Gigabit per second with a comparatively low delay and gives way
for connecting 100 billion plus devices. These technologies are still in their infancy
and are presently being tested in a number of pilot projects. It is anticipated that 5G
networks will be widely deployed by 2020. Fast and reliable Internet access as well
as support in order to construct smart cities will be made possible by the introduc­
tion of 5G networks. New designs, including cloud RAN and virtual RAN, will be
introduced by these networks, which can permit most server farms via localized data
centers at the network edges, and increasingly centralized network setup [17].

Network function virtualization: Network service providers have been worried

about managing network infrastructure. This issue hampers innovation in the telecom
136 IoT and Machine Learning for Smart Applications

sector and impacts income, too. In order to reduce or completely eliminate their reli­
ance on proprietary hardware, network operators. Network function virtualization
emerged from the need to change the paradigm away from relying solely on hardware
to perform key network functions (Network Functions Virtualization [NFV]) [18].
Industrial NFV specifications were created by a collaboration of telecom carriers from
the European Telecommunications Standards Institute. NFV is a technique developed
to benefit from the development of virtualization in information technology. Utilizing
commercially available hardware, this technology converts network functions that are
hardware-based into software-based applications.
A few benefits of this emerging technology include platform transparency, expan­
sion, adaptability, enhancements in efficiency, lower capital, and operational expenses.
The following elements are supported by NFV: physical servers, hypervisors, and
guest virtual machines. Physical resources like RAM, storage space, and CPU are
offered by the physical server. The environment in which a guest virtual machine can
function is created by a virtual machine monitor known as a hypervisor [18]. The
necessary software is started on the guest virtual machine, a piece of software that
mimics the appearance and functionalities of an actual platform.

Quantum networking: In some respects, we were aware of a class of issues that

a computer would never be able to handle even before the first actual computer. In
1994, Peter Shor accomplished a similar accomplishment by creating an algorithm
for a yet-to-be-built quantum computer that can do prime factorization in a finite
amount of time. Long before any such machine had been constructed, the potential
(and constraints) of this unique sort of mechanical processing were being outlined. In
the field of computing, quantum computers are a new, potentially revolutionary tech­
nology [19]. Quantum Networking, a related upcoming technology, allows quantum
bits (qubits) to be transmitted between quantum networks. Like many others, there is
a lack of any specific knowledge on whether quantum networking will be an obscure
detour in the development of digital networks or whether it will establish the standard
mainstream basis for tomorrow’s digital services.
Quantum networking utilizes principles of quantum mechanics to enable secure
and efficient communication and information processing. Quantum networking
involves the development of technologies and protocols for the secure transmission
of quantum information over long distances. It leverages quantum phenomena such
as entanglement and superposition to achieve tasks such as quantum key distribution
(QKD), quantum teleportation, and distributed quantum computing. Applications
of quantum networking include quantum cryptography for secure communication,
distributed quantum computing for solving complex problems, quantum sensor
networks for high-precision measurements, and quantum-enhanced machine learning
for accelerated data analysis.

8.4 BIG DATA IMPLEMENTATIONS IN IOT-ENABLED CITY OF

THE FUTURE
Cities all across the world have adopted the buzzword “smart city,” which is gradually
emerging as a dominant concept for urbanization. Big data technologies are used in
Realization of Smart City Based on IoT and AI 137

smart cities to populate data and enhance many solutions for smart urban areas. Big
data can be used to build any resources or services needed for a smart city [20]. Big
Data requires the appropriate concepts for effective data analysis in order to attain its
objectives and enhance utilities in smart urban areas. These methods and tools may
promote interaction and collaboration among entities, offer assistance to many indus­
tries in the smart city, and enhance customer perspectives and investment prospects.

• Smart power grid

This idea is inextricably linked to the emergence of sustainable sources that are ever­
evolving and reducing dependency on fossil fuel substitutes while maintaining equi­
librium between supply and demand [20]. A typical component to spot in this is a
dual-directional energy flow, which suggests that users are no longer only passive
consumers but also energy producers [21]. The main challenge for the smart grid is
the networks themselves, as the majority of them were built to serve massive plants
that run on fossil fuels, supplying power to customers only in a single way [22]. Three
essential elements are required for this bidirectional flow: storage technology, man­
agement systems, and infrastructure supervision. Technical assistance is encouraged
by each of these elements.
Studies have been trying to incorporate, assess, and use actual information on
electricity generation and demand, as well as the types of climatic data due to this
rapid spread of smart grids. Significant investment effectiveness of the current smart
grid infrastructure is anticipated as a result of the improved system performance and
various performance metrics. A significant amount of data is produced in this envir­
onment by varied sources like user energy consumption patterns, phasor measure­
ment data for insight, and power usage data measured by widely used smart meters
between multiple other sources [23].
Big data gathered from the smart grid environment used effectively will aid to take
decision in making wise decisions regarding the quantity of electricity to be offered
along with meeting user demand. Moreover, the analysis of the smart grid data can
help to achieve well-planned objectives by providing a particular assessment plan that
is reliable as per the source, request, and functional prototypes. The data analysis of
these data could be used to foresee the future demand for power supply [24].

• Smart healthcare

The concept of “Smart Planet,” first presented by International Business Machines

Corporation (IBM), sparked the creation of smart healthcare. This concept is an intel­
ligent system that collects data from sensing devices, transfers it over the IoT, and
then uses cloud computing and processing power to process it [25]. In terms of man­
aging global civilization in an interactive manner, it can synchronize and incorporate
social systems. “Smart healthcare” refers to healthcare that makes use of the latest
trends like wearables, the IoT, and mobile internet to gather data vigorously, inter­
link various healthcare-related components and organizations, and then vigorously
accomplish as well as intelligently react to the requests of the health eco-system.
138 IoT and Machine Learning for Smart Applications

IoT-enabled medical care has the potential to facilitate communication among all
involved parties in the healthcare sector, guarantee that participating individuals
obtain the services they need, support decision-making, and encourage the effective
use of technology. Smart healthcare is about raising the bar for information creation
within the medical system.
The medical field has produced a vast amount of material over the years. The
world’s population increase has enabled abrupt changes in diagnosis deploy­
ment methods, with various conclusions driving these changes being data-driven.
Healthcare professionals possess the ability to collect and assess medical data that
can help administrative organizations and health insurers by using the appropriate
analytics technologies. Also, efficient analytics of large medical data sets can improve
quality of life, prevent unnecessary deaths, and help predict the emergence and spread
of treatment options and diseases. Smart devices that are linked to homes or clinics
to keep a check on patients, might raise the volume and consistency of informa­
tion collected for certain individual’s health conditions. This information is used to
interpret clinical notes. Additionally, the analysis of vast volumes of healthcare data
enables physicians to recognize warning symptoms of a severe illness during the ini­
tial therapy phases, potentially saving many patients [26].

• Smart Transportation

The growth of smart transportation can be impacted by transportation governance.

Smart transportation governance in smart cities consists of a number of plans,
regulations, strategies, and projects. Automated ticketing, electric automobiles,
automated vehicles, and clean transportation legislation are a few examples of these
approaches. In the transportation industry, the term “smart” can refer to new forms
of propulsion (such as electricity), improved vehicle controls (such as the Intelligent
Transport System), improved business models (such as car sharing), improved
regulations, and improved transport planning and policy. Their primary goals are to
reduce pollution, transportation congestion, increase safety, accelerate transfers, and
lower travel expenses [27].
In order to reduce traffic congestion by offering alternate routes and lowering acci­
dent rates by evaluating the mistakes in the past, along with aspects like finding the
reason for the mistake and driving speed, the patterns discovered out of the immense
quantities of traffic data can aid the system of commuting. Transport systems’ data can
also be used to optimize freight movements [28]. The large amounts of data collected
by smart transportation systems can also help with cargo optimization and consolida­
tion by reducing supply chain waste. The data gathered from the IoT-enabled trans­
portation system can also offer various advantages, including lowering environmental
impact, boosting safety, and enhancing end-to-end customer experience.

• Smart Governance

Governments are having to reevaluate their obligations in a knowledge-based society

as a result of the expanding role that technologies play in the operation of muni­
cipal systems. Prior studies have called this capacity “smart governance” [29].
Realization of Smart City Based on IoT and AI 139

Despite its significance, there isn’t a consensus on what this idea means. Many have
concentrated on the procedure for gathering various types of data as well as info
pertaining to public administration by sensors or sensor networks [29] or on the
accomplishment of the social integration of city dwellers using government ser­
vices, whereas some earlier research focuses on both political engagement and the
government’s operation [30].
Smart governance can be enabled by big data analytics [31]. Analysis of data
that can lead to collaborations between them can quickly identify organizations or
agencies with similar goals. Collaboration like this can help countries develop. Since
governments are well-versed in the requirements of individuals concerning healthcare
and societal well-being, training, and other sectors, big data analytics may also help
them create and execute effective policies. A review of comprehensive data from mul­
tiple academic institutions may also help to reduce unemployment.

8.5 TECHNICAL CHALLENGES

The increasing need for big data and smart cities fosters innovation and highlights
the importance of developing new smart apps. However, good data management is
required in order to enhance the smart city’s services. These technological issues
with big data and smart cities are briefed in this subsection. It takes the integration
of several technologies to create a smart city. The adoption of smart city technology
is further hampered by the QoS given by various technologies [32]. For instance, the
aim to fulfill core objectives like fault-tolerant, scalable, and dependable networks
cannot be sacrificed. Similarly, it is challenging to design massively scalable data
analysis and storage systems supported by wise cloud service choices. Before the
urban planning application is fully integrated, the QoS offered by these technologies
must be adequate. The approaches and frameworks for establishing and implementing
QoS criteria in a smart city are crucial [33]. Artificial intelligence is a subset of know­
ledge engineering that makes use of computational intelligence concepts like neural
networks, genetic algorithms, and bioinformatics. These algorithms are effective, effi­
cient, and robust [34]. However, small data sets are the only size for which machine
learning algorithms are effective, efficient, and robust [35]. As a consequence, data
analytics for smart cities cannot use these algorithms. Because smart cities generate
large volumes of data, conventional computational intelligence approaches are no
longer applicable to large data analytics.
With the aid of a variety of embedded intelligent units, data from smart cities may
be gathered in many different ways. Data aggregation among these cities represents
one of the greatest difficulties that must be overcome, though, as the target of the
smart-city drive is to mix massive data generated within various origins [36]. Smart
cities have included a number of technologies in recent years, lowering the techno­
logical obstacles to handling the data. However, one of the toughest issues with any
data combining system is quality data, particularly if the information is unreliable,
incomplete, inconsistent format, or insufficient [37]. Data analysis is regarded as an
essential tactic for fostering progress and welfare in every modern city. To enhance
the life quality for inhabitants & create reliable cities, processing issues with this data
140 IoT and Machine Learning for Smart Applications

must be resolved [38]. In a smart city, data is extracted from numerous sources. To
interpret the data and make decisions, one needs special algorithms and visualization
techniques, which have an impact on activities related to smart cities. For instance,
synchronizing the consumption detected by consumers’ meters with that estimated
by other utilities’ systems helps reduce energy/water losses due to defective systems
[39]. In the big data era, personal data in smart cities is susceptible to analytics, trade,
and misuse, which creates worries about control loss, theft, and profiling [39]. People­
specific information about residents, such as social events and places, is gathered
daily, for instance. The massive amounts of personal data that smart city technology
has collected are being protected against theft and hacking despite countless attempts
to do so. A number of cyber-security issues related to smart city technology need to
be addressed, even though city-level successful cyberattacks are still comparatively
infrequent.

8.6 CONCLUSION
This chapter delves into the burgeoning proliferation of interconnected devices within
urban landscapes, capturing the interest of scholars across diverse disciplines. The
central focus is a comprehensive exploration of the utilization of big data analysis
within the context of smart cities. The genesis of the “smart cities” idea is intrin­
sically connected to the remarkable expansion of the IoT, catalyzing the evolution
of urban infrastructures toward intelligence. Despite this paradigm shift, the matur­
ation of smart cities is still in its nascent phases, necessitating continuous innovation,
robust networking frameworks, and adept data management capabilities. Establishing
a pragmatic foundation for smart cities holds significant esteem among both aca­
demic circles and industry practitioners. This analysis elucidates pivotal technolo­
gies integral to intelligent urban frameworks, offering insights into specific smart
city initiatives that stand to benefit significantly from the application of big data
analytics. Additionally, prospective business models and architectural intricacies are
scrutinized, with a distinct emphasis on the efficacious handling of voluminous data
sets. In summation, the discourse underscores the pivotal role of extensive data in
facilitating informed decision-making and extracting valuable knowledge within the
ambit of smart city development.

REFERENCES
[1] Nathali Silva, Bhagya, Murad Khan, et al. “Big data analytics embedded smart city
architecture for performance enhancement through real-time data processing and
decision-making”. Wireless Communications and Mobile Computing, no. 1 (2017):
9429676.
[2] Susmitha, K., & S. Jayaprada. “Smart cities using big data analytics”. International
Research Journal ofEngineering and Technology (IRJET) 4, no. 8 (2017): 1615-1620.
[3] Hashem, Ibrahim Abaker Targio, Victor Chang, et al. “The role of big data in smart
city”. International Journal of Information Management 36, no. 5 (2016): 748-758.
[4] Silva, Bhagya Nathali, Murad Khan, Changsu Jung, & Jihun Seo. “Urban planning and
smart city decision management empowered by real-time data processing using big
data analytics”. Sensors 18, no. 9 (2018): 2994.
Realization of Smart City Based on IoT and AI 141

[5] Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. “Internet of Things (IoT): A
vision, architectural elements, and future directions”. Future Generation Computer
Systems 29, no. 7 (2013): 1645-1660.
[6] Hashem, I. A. T, Yaqoob, Anuar, N. B., Mokhtar, S., Gani, A., & Khan, S. U. “The
rise of big data on cloud computing: review and open research issues”. Information
Systems 47 (2015): 98-115. http://dx.doi.org/10.1016/j.is.2014.07.006.
[7] Hashem, I. A. T., Anuar, N. B., Gani, A., Yaqoob, I., Xia, F., & Khan, S. U.
“MapReduce: Review and open challenges”. Sciento Metrics, (2016): 1-34. http://
dx.doi.org/10.1007/s11192-016-1945-y.
[8] Mell, Peter. “The NIST Definition of Cloud Computing.” NIST Special
Publication (2011): 1-7.
[9] Chang, V., David, B., Wills, G., & De Roure, D. “A categorisation of cloud business
models.” In CCGrid, 10th International Symposium on Cluster, Cloud and Grid
Computing. 2010.
[10] Armbrust, M., Fox, A., Griffith, R., Joseph, A. D., Katz, R., Konwinski, A., & Stoica,
I. “A view of cloud computing”. Communications of the ACM 53, no. 4 (2010): 50-58.
[11] Atzori, L., Iera, A., & Morabito, G. “The internet of things: A survey”. Computer
Networks 54, no. 15 (2010): 2787-2805.
[12] Kang, Y.-S., Park, I.-H., Rhee, J., & Lee, Y.-H., “MongoDB-Based repository design
for IoT-generated RFID/sensor big data”. IEEE Sensors Journal 16, no. 2 (2016): 485­
497. http://dx.doi.org/10.1109/jsen.2015.2483499 .
[13] Dargie, W. W., & Poellabauer, C., “Fundamentals of Wireless Sensor Networks: Theory
and Practice”. John Wiley & Sons (2010).
[14] Al Nuaimi, E., Al Neyadi, H., Mohamed, N., & Al-Jaroodi, J., “Applications of big data
to smart cities”. Journal of Internet Services and Applications 6, no. 1 (2015): 1-15.
[15] Oualhaj, O. A., Kobbane, A., Sabir, E., Ben-othman, J., & Erradi, M., “A ferry-assisted
solution for forwarding function in Wireless Sensor Networks”. Pervasive and Mobile
Computing 22 (2015): 126-135. http://dx.doi.org/10.1016/j.pmcj.2015.05.003Owen.
[16] Lee, J.-S., Su, Y.-W., & Shen, C.-C., “A comparative study of wireless
protocols: Bluetooth, UWB, ZigBee, and Wi-Fi”. Paper presented at the 33rd Annual
Conference of the IEEE Industrial Electronics Society, 2007. IECON (2007).
[17] Abdalla, I., & Venkatesan, S., “Remote subscription management of M2Mterminals in
4G cellular wireless networks”. Proceedings of the 37th Annual IEEE Conference on
Local Computer Networks (LCN 2012): 869-877.
[18] Hawilo, H., Shami, A., Mirahmadi, M., & Asal, R. “NFV: State of the art, challenges,
and implementation in next generation mobile networks (vEPC)”. IEEE Network 28,
no. 6 (2014): 18-26.
[19] P. W. Shor, “Algorithms for quantum computation: discrete logarithms and factoring”.
Proceedings 35th Annual Symposium on Foundations of Computer Science
(1994): 124-134. http://dx.doi.org/10.1109/SFCS.1994.365700.
[20] Crossley, P., & Beviz, A. “Smart energy systems: Transitioning renewables onto the
grid”. Renewable Energy Focus 11 (2010): 54-59.
[21] Lund, H., Andersen, A. N., 0stergaard, P. A., Mathiesen, B.V., & Connolly, D. “From
electricity smart grids to smart energy systems—A market operation based approach
and understanding”. Energy 42 (2012): 96-102.
[22] “Renewables Won’t Save Us If the Electric Grid Is Not Ready”. Available
online: www.forbes.com/sites/davidblackmon/ 2020/09/30/renewables-wont-save-us-
if- the-electric-grid-is-not-ready/?sh=76da985c7abf (accessed May 23, 2022).
[23] Zhang, Y., Huang, T., & Bompard, E.F., “Big data analytics in smart grids: a review”.
Energy Informatics, 1, no 1 (2018): 1-24.
142 IoT and Machine Learning for Smart Applications

[24] Al Nuaimi, E., Al Neyadi, H., Mohamed, N., & Al-Jaroodi, J. “Applications of bigdata
to smart cities”. Journal of Internet Services and Applications 6, no. 1 (2015): 1-15.
[25] J. L. Martin, H. Varilly, J. Cohn, & G. R. Wightwick, “Preface: Technologies for a
smarter planet”. IBM Journal of Research and Development, 54, no. 4 (July-Aug,
2010): 1-2. http://dx.doi.org/10.1147/JRD.2010.2051498 .
[26] Roy, N., Pallapa, G., & Das, S.K. “A middleware framework for ambiguous context
mediation in smart healthcare application”. Third IEEE International Conference on
Wireless and Mobile Computing, Networking and Communications, 2007. WiMOB
(2007).
[27] Benevolo, C., Dameri, R. P., & D’Auria, B. “Smart mobility in smart city”. In: Torre,
T., Braccini, A., Spinelli, R. (eds.) Empowering Organizations. Lecture Notes in
Information Systems and Organisation, 11. Cham: Springer (2016). https://doi.org/
10.1007/978-3- 319-23784-8_2.
[28] Ju, G., Cheng, M., Xiao, M., Xu, J., Pan, K., Wang, X., & Shi, F. “Smart transportation
between three phases through a stimulus-responsive functionally cooperating device”.
Advanced Materials 25, no. 21 (2013): 2915-2919.
[29] Giffinger, Rudolf, Christian Fertner, Hans Kramar, Robert Kalasek, Natasa Pichler-
Milanovic, & Evert J. Meijers. “Smart cities. Ranking of European medium-sized
cities. Final Report.” (2007).
[30] Schuurman, D., Baccarne, B., De Marez, L., & Mechant, P. “Smart ideas for smart
cities: Investigating crowdsourcing for generating and selecting ideas for ICT innov­
ation in a city context”. Journal of Theoretical and Applied Electronic Commerce
Research 7, no. 3 (2012): 49-62.
[31] Caragliu, A., Del Bo, C., & Nijkamp, P. “Smart Cities in Europe”. Proceedings to
the 3rd Central European Conference on Regional Science, Kosice, Slovak Republic,
45-59 (2009).
[32] Bellavista, Paolo, Antonio Corradi, & Andrea Reale. “Quality-of-service in data center
stream processing for smart city applications.” In: Samee U. Khan, Albert Y. Zomaya,
& Elias Houstis. Cham (eds.), Handbook on Data Centers, 1047-1076. Cham,
Switzerland: Springer (2015).
[33] Jalali, R., El-Khatib, K., & McGregor, C. (2015). “Smart city architecture for commu­
nity level services through the internet of things”. Paper Presented at the 2015 18th
International Conference on Intelligence in Next Generation Networks(ICIN).
[34] Jin, J., Gubbi, J., Marusic, S., & Palaniswami, M. “An information framework for cre­
ating a smart city through internet of things”. IEEE Internet of Things Journal 1, no. 2
(2014): 112-121.
[35] Tsai, C.-W., Lai, C.-F., & Vasilakos, A. V. “Future Internet of Things: open issues and
challenges”. Wireless Networks 20, no. 8 (2014): 2201-2217.
[36] Su, K., Li, J., & Fu, H. (2011). “Smart city and the applications”. International
Conference on Electronics, Communications and Control (ICECC) (2011).
[37] Gouveia, J. P., Seixas, J., & Giannakidis, G. “Smart city energy planning integrating
data and tools”. Proceedings of the 25thInternational Conference Companion on
World Wide Web (2016).
[38] Gulisano, Vincenzo, Magnus Almgren, & Marina Papatriantafilou. “When smart cities
meet big data.” Smart Cities 1, no. 98 (2014): 40.
[39] Tene, O., & Polonetsky, J. “Privacy in the age of big data: A time for big decisions”.
Stanford Law Review Online, 64 (2012): 63.
 Sentiment Analysis
9 of Airline Tweets Using
Machine Learning
Algorithms and Regular
Expression
S. Nagendra Prabhu, A.P. Rohith,
Shubhankar Bhope, and P. Sivakumar

9.1 INTRODUCTION
Sentiment analysis, as the name implies, is determining the perspective or feeling
underlying an event. Essentially, it entails analyzing and determining the emotions
and intents underlying text, speech, and modes of communication. We, humans, com­
municate in a variety of languages, but each language is merely an intermediary or
means through which we seek to express ourselves. And we are attached to every­
thing we say. It can be positive, negative, or neutral. Suppose you have a restaurant
chain that sells a variety of foods such as dosa, paratha, naan, and milkshakes. You
created a website to sell groceries. Clienteles may now order items from your website
and even leave reviews.

User Review 1: I love this butter naan; it’s so delicious.

User Review 2: This chicken dosa has a very bad taste.
User Review 3: I ordered this paratha today.

The first review was positive, and that means the customer was really satisfied with
the butter naan. The second opinion is negative; therefore, the company must review
its dosa division. The third does not mean whether the customer is satisfied, so we
can take that as a neutral statement. However, now a problem arises: there will be
hundreds and thousands of user reviews for their product, and after a while, reviewing
each user review and making conclusions will become almost impossible. They also
can’t come to a conclusion by taking only about 100 reviews because maybe the first
100-200 customers tasted the same and liked the naan, but over time, they didn’t.
With rating increases, there can be a situation where none of the positive reviews can
cross negative reviews.

DOI: 10.1201/9781032623276-9 143

144 IoT and Machine Learning for Smart Applications

©O¥ • v
FIGURE 9.1 Emoji for sentiment analysis.

9.2 TEXT SENTIMENT CLASSIFICATION

People love to express their feelings. Happy or unhappy. Like or dislike. Compliment
or complaint. Good or bad, as shown in Figure 9.1. That is, positive or negative.
Emotion analysis in natural language processing involves decoding these emotions
from the text. Is it positive, negative, or both? If there is a sentiment, what object in
the text does that sentiment refer to, and actual emotional phrases such as poor, lack­
luster, cheap, etc. (not only positive or negative)? As a technique, sentiment analysis
is both interesting and useful. First, the enjoyable part. It’s not always clear, at least
for a computer program, if a text’s emotion is positive, negative, or mixed. The signals
can be subtle. Common sense aside, it’s even harder to discern which objects in the
text are the subjects of which feelings, especially when both positive and negative
feelings are involved. Emotion analysis focuses on a writing’s polarities (good - posi­
tive, bad - negative, or neutral). However, it appears that divergence can distinguish
individual thoughts and sentiments (angry, pleased, sad, etc.), urgency (essential, not
important), and even purpose. You can create and modify classifications to meet your
sentiment evaluation requirements, contingent on how you want to interpret con­
sumer feedback and queries.

9.3 LITERATURE INVESTIGATION - GENERAL

This chapter affords an exhaustive explanation of the frequent processes and strat­
egies we looked into and used to comprehend the approach necessary for our project.
In order to maximize the speed, efficiency, and quality of classification and detection
in our system, we improved our algorithm with the help of the reference papers we
studied. These improvements allowed us to explore effective solutions to some of the
problems that were present in the classification of text.

9.3.1 Current System

A literature study looks for systems that are comparable to existing ones and
determines how the researcher’s project differs from them. This makes it easier to
understand the project in depth. It gives a subject-specific blend of theoretical, meth­
odological, and current understanding of findings. Information must be gathered in
unity with the project. The following part outlines our model and how it differs from
other similar models.
A study by Schaffer and Huynh [1] conducted a pictorial proportional investigation
of the various progressions to investigate the characteristics of the course assembly
variations by enchanting into justification the students’ behavioral characteristics. It
Sentiment Analysis of Airline Tweets Using Machine Learning 145

produced several endorsements for instructors and course designers. In their research
[2], Kevin Andrey and Liedy offered three methods for categorizing the moods of
tweets. A strategy centered on word, emotion, and a hybrid approach. The sentiment
infused into the emotions is taken into consideration as a criterion for categorizing
the messages in the emotion-based method. The inclusion of an emoticon was a
selection factor for the tweet. Words like bad, excellent, good, etc., that reflect sen­
timent are engaged in interpretation when using the word-based technique to infer
sentiment. In the hybrid method, the use of words and emoticons was taken into
consideration. Stoffova et al. [3] executed the three main experiments to determine
the student’s place of origin in relation to technical awareness in a global educational
setting. Students from two nations on two distinct continents took part in this study.
They used principal component analysis (PCA) and support vector machine (SVM)
to do the same. Adarsh and Ravikumar [4] performed a similar study. By taking into
consideration the tweets from three well-known Airlines, a performed efficient yet
straightforward way of recognizing feelings on Twitter is proposed. Based on the
score calculation, sentiments were classified as negative, neutral, or positive. Torre-
D^ez et al. [5] castoff two machine learning processes to perform the basic architec­
ture. The process involves a data gathering phase, preliminary text processing phases
over tumbling incorrect information, and expression-entrenching strategies for reno­
vating text-based information into numerical-constructed information; Convolutional
Neural Network’s (CNN’s) algorithm can automatically determine qualities.
Feedforward Neural Network (FFNN) for manipulative neither positive sentiment
score (PSS) nor negative sentiment score (NSS) requirements, and Mamdani Fuzzy
System (MFS) for categorizing the data they provide as favorable, unfavorable, or
neutral. Ishaq et al. [6] provide a sentiment evaluation categorization strategy that
employs a CNN’s besides a GN (genetic algorithm) strategy. After mining semantic
structures, many techniques, including the recommended CNN-based ensemble,
Support Vector Machine, random forests, maximum entropy, stable differential inves­
tigation, decision tree, and generalized linear conventional, are trained concurrently.
Lei et al. [7] propose an algorithm for suggestions based on sentiment data extracted
from user feedback on social media platforms. To achieve the assessment prediction
goal, they combine user opinion, resemblance interaction sentiments influence, and
element reputation similarity in a solitary matrix factoring method. They mainly rely
on social media users’ emotional responses to determine their preferences. Chen et al.
[8] take an alternative strategy. A cross-modal hypergraph prototype was developed to
combine documentary and emotive data for sentiment categorization, which is useful
for producing online service predictions. Furthermore, the Latent Dirichlet Allocation
(LDA) concept prototypes have been implemented into our projected cross-modal
hypergraph framework, allowing us to better interpret global higher-level facts
and eliminate ambiguity in certain specific phrases. Manshu and Bing presented a
Hierarchical Attention Network with Priority (HANP) method for Context-Dependent
Sentiment Classification (CDSC) employment [9]. When classifying sentiment, the
recommended HANP can prioritize keywords and phrases. The sentiment vocabu­
lary matching stage can identify key pivots and non-pivots. By learning synonyms,
the HANP can identify and understand terms related to both pivot and non-pivot (or
dis-pivot) concepts. Abdalgader and Shibli [10] presented a novel application of the
146 IoT and Machine Learning for Smart Applications

lexicon-based keyword polarities detection approach to various customer evaluations

based on the convention of contextual expansion. The modified methodology finds a
word’s polarities by examining the semantic relationships among its context enlarge­
ment, equivalents, and glosses.
Lee [11] studies the early effect of COVID-19 emotion on the United States
stock market by consuming big data statistics, namely the day-to-day news emotion
Index Daily New’s Mood Index (DNSI) and Google’s Tendencies information on
coronavirus-connected search terms. The goalmouth is to explore links between
COVID-19 emotions and 11 other US market segment directories over a certain
period. Yu et al. [12] proposed a Chinese Emotional - Sentimentality evaluation
approach that utilized a larger dictionary. The extended sentiment dictionary consists
of normal sentiment dictionaries, field sentiment sentences, and polysemic field
emotion keywords. The Naive Bayesian field classifier determines the sentimentality
polarity of a polysemic sentiment word within a text. Rathi et al. [13] suggested
an article to discuss. To assess the classifier’s overall precision in tweet categoriza­
tion, they employed SVM, AdaBoosted Decision Tree, and Decision Tree constructed
hybrid emotion cataloguing methods. Our suggested methodology classifies tweets
as detrimental or favorable and then employs sentiment analysis to make subsequent
decisions. We employ preprocessing techniques to offer accurate data as input for the
training procedure. Zhong et al. [14] introduced stock-NF, a flow-oriented generating
framework aimed at forecasting stock undertaking. To aid learning in being more
adaptable and evocative, latent depictions of tweets besides antique stock values, the
proposed method makes use of normalizing flows, which are extensively employed in
the creation of pictures and audio synthesizing. Cheng and Yue [15] introduced Part-
of-Speech and Targeted Attention Network (POS-TAN), a sentiment identification
model they developed. The framework comprises the POS-attention mechanisms,
which accumulate emotive content embedded in parts of speech, and the self-attention
method, which is used to understand the text’s feature expression. Furthermore, they
employ the focal loss to mitigate the effects of sampling instability on the organiza­
tion effect. Table 9.1 shows the literature survey of the sentiment analysis model.

9.4 LITERATURE REVIEW - GENERAL

• The most frequently used procedures for the classification of text are SVM,
linear regression, and Naive Bayes.
• Multinomial NB is most effective for two-class issues, which presents a bar­
rier when using it. This possibly will inhibit our effort because of the degree
of the dataset.

The proposed method categorizes text based on sentiment using a Random Forest­
based Classifier. The evaluation of a considerable number of research articles found
that the following algorithms are most commonly utilized for sentiment analysis in
text: support vector machine and random forest.
The challenge of using the support vector machine is that it is superlative suited
for two-class complications. The size of the dataset may have an impact on our work.
 Sentiment Analysis of Airline Tweets Using Machine Learning
TABLE 9.1
Literature Survey

Algorithm
Publication Details Methodology Used Advantages Limitations
I. D. L. TorreDrez. The basic architectural hybrid model was carried Deep learning Predicting highest • Time ingesting is
et al.. "Sentence Level out by I. D. L. Torre Dez. et al., data collection classifier accuracy 90% with more.
Classification Using phase, the preprocessing of texts processes to using parallel this model. • Space cast-off by a
Parallel Fuzzy Deep reduce noisy data. Methods for transforming fuzzy. model is very large.
Learning Classifier". IEEE text-based data into numeric representations
Access, vol. 9. pp. 17943- using word embedding techniques.
17985. 2021. Convolutional Neural Network (CNN) methods
for obtaining characteristics immediately.
FFNN calculates PSS and NSS values, while
MFS classifies the information it receives into
neutral, negative, or positive classifications.
V. S toffova. Z. illes et al.. The three main tests were carried [3] to Support vector Clarity of data and ease • Limited number
"Machine Learning-Based determine the student's country of origin machine. of training. of extraction
Student's Native Place with regard to technical knowledge in a PCA. techniques.
Identification for Real- global educational setting. The participants • Overfitting, data
Time". IEEE Access, vol. in this research study were students from two leakage.
8. pp. 130840-130854. nations located on opposite continents. They
2020. accomplished the same using PCA and SVM.
Brandon Huynh and James When making predictions, it seems to be conducted TF-IDF vector In this particular' case Course structure was
Schaffer. "An Analysis a pictorial proportional examination of the space models. study, employing a considered when
of Student Behavior in various progressions to examine the effects of visual and systematic making predictions.
Two Massive open online the variations in development assembly while analytics approach The predictive power
courses". IEEE/AC M captivating into interpretation the behavioral proved effective in of the algorithm was
ASONAM. 2016. characteristics of the students. They then elucidating student not good.
provided a number of recommendations for achievement.

147
teacher's and course designer's.
(continued')
 148
TABLE 9.1 (Continued)
Literature Survey
Algorithm
Publication Details Methodology Used Advantages Limitations
T. Zhong, et al.. "Learning This article offered StockNF. a flow-based StockNF Improved the/- Overfitting, data
Sentimental and Financial framework for forecasting fluctuations measure of sentiment leakage.
Signals With Normalizing in stocks. The proposed methodology prediction and overall
Flows for Stock Movement employs flow normalization to empower the classification accuracy.
Prediction". IEEE Signals- development of more accommodating and
Processing Letters, vol. 29. communicative latent illustrations for tweets
pp. 414-418. 2022. and previous stock values.

loT and Machine Learning for Smart Applications

M. J. Adarsh and By analyzing tweets from three well-known Score We were easily able The problem of this
P. Ravikumar. "An airlines, this article suggests an efficient calculation to find difference method is that it
Effective Method of yet straightforward method for identifying Score = No. of between positive and may not achieve
Predicting the Polarity sentiment on Twitter. The calculation of positive words negative sentiment of the desired results
of Airline Tweets using the score served as the basis for identifying - No. of negative twitter users. when applied to
sentimental Analysis". positive, negative, and neutral moods. words. mocking twitters
2018 4th ICEES. 2018. since the placement
of optimistic and
undesirable results
leads to differing
conclusions.
Liedy's del Carmen et al.. The methodology utilized was CRISP, which CRISP-data This technology saves Overfitting and
"Analysis of Behavioral involved evaluating descriptive models mining money since it employs increment in size of
of Customer in the Social through clustering and association rules. techniques. a number of methods training datasets.
Networks using Data The findings indicate that the proposed that eliminate easy
Mining Techniques". IEEE models can offer valuable insights for data mining activities
/ACMASONAM. 2016. crafting marketing strategies tailored to user and are widely used in
preferences. business.
 Sentiment Analysis of Airline Tweets Using Machine Learning
Lee. H. S. "Leveraging big Lee et al., examines the initial influence of Time series The study identifies Overfitting
data to investigate the COVID-19 sentiment on the US stock market regression industry-specific
initial effects of COVID 19 by leveraging big data, specifically the Daily models. COVID-19 attitudes
mood on the United New's Mood Index (DNSI) and Google and categorizes them
State stock market". Trends data pertaining to coronavirus-related based on correlation.
Sustainability, vol. 12. searches. The aim is to explore the correlation
no. 16. p. 6648. 2020. between COVID-19 sentiment and 11 distinct
categories within the US stock market
directories over a specified time frame.
A. Ishaq et al.. "Aspect Based A highly effective classification performance CNN and GA. Parallelism. Global Being stochastic, there
Sentiment Analysis Using for sentiment analysis can be achieved using optimization. are no guarantees
a Hybridized Approach a CNN combined with a genetic algorithm. on the optimality
Based on CNN and GA". Additionally. Support Vector Machine. CNN­ or the quality of the
IEEE Access, vol. 8. 2020. based ensemble methods. Random Forest. solution.
Entropy. Stabilized Discriminant Analysis.
Decision Trees, and various other models are
trained following the extraction of semantic
data.
G. Zhao et al.. "Rating This article proposes a recommendation RPS model. The findings False positive is very
Prediction Based on mechanism based on sentiment data context-MF. confirmation that high.
Social Sentiment From gathered from user evaluations on social PRM. and the three sentimental
Textual Reviews." broadcasting. To solve the rating prediction EFM. components
IEEE Transactions on problem, they incorporate handler sentiment significantly influence
Multimedia, vol. 18. no. 9. resemblance, interpersonal sentiment effect, the prediction of
pp. 1910-1921. Sept. 2016. and item character similarity in a solitary rating. On a real-
matrix factoring approach. They specifically world dataset, it also
exploit user sentiment on social media to demonstrates notable
demonstrate preferences. gains over existing
techniques.

149
(continued)
 150
TABLE 9.1 (Continued)
Literature Survey
Algorithm
Publication Details Methodology Used Advantages Limitations
T. Manshu et al.. "Adding They propose using the HANP technique to CNN and RNN. The effectiveness of Overfitting
Prior Knowledge in the CDSC problem in this chapter. When HANP has been tested
Hierarchical Attention recognizing sentiment, the recommended using the Amazon
Neural Network for HANP can focus on significant words and review dataset. We
Cross Domain Sentiment phrases. The emotion dictionary match layer achieve cutting-edge
Classification". IEEE tracks key pivots, non-pivots. and dis-pivots. accuracy with this
Access, vol. 7. pp. 32578- By understanding its equivalents, the HANP dataset.

loT and Machine Learning for Smart Applications

32588. 2019. can figure out what a non-pivots or dis-pivots
is.
Z. Chen et al.. To enhance emotional categorization for Cross-modal The theme-created • High time
"TCMHG: Topic-Based online services, this article proposes the hypergraph mixture model can complexity
Cross-Modal development of a cross-modal hypergraph model. SVM. improve classification Overfitting
Hypergraph Learning model that integrates textual and emotional LDA topic performance even
for Online Service information separately. Additionally, the LDA model. further and cut down
Recommendations". IEEE (Latent Dirichlet Allocation) topic model on computational
Access, vol. 6. pp. 24856- is integrated into the proposed cross-modal expenses.
24865. 2018. hypergraph model to incorporate higher-level
global knowledge and to reduce ambiguity in
certain phrases.
 Sentiment Analysis of Airline Tweets Using Machine Learning
A. Shibli et al.. This article presents a revised version of the Lexical analysis. It was able to detect • Less accuracy
"Experimental Results on lexicon-based word polarity identification negation scope of the • 78%
Customer Reviews Using approach that uses context expansion in word. • Accuracy
Lexicon-Based Word response to several user critiques. The
Polarity Identification improved method computes the semantic
Method". IEEE Access, relatedness of context expansions identified
vol. 8. pp. 179955-179969. in WordNet synonyms and glosses in order to
2020. control the polarity of an estimated term.
A. Malik et al.. "Sentiment They suggest the POS-TAN sentiment SVM. Improved the F- • Disregards the
Analysis of Tweets categorization model in this chapter. This Adaboosted measure of sentiment aspect-related data.
Using Machine Learning model comprises the POS attention, which is Decision prediction and overall A sentence may
Approach". 2018 11th used to collect emotive information contained tree and decision classification accuracy. have numerous
International Conference in parts of speech, in addition to the self­ tree. diverse features, and
on Contemporary attention mechanism, which is used to learn it is occasionally
Computing (11th ICCC). the text's feature expression. In order to important to
pp. 1-3. 2018. lessen the effect of sample imbalances on the determine the
classification effect, they also introduce focus sentiment polarity of
loss. a particular aspect.
F. Li et al.. "Chinese Text In this article, a more sophisticated dictionary­ Naive Bayesian This approach guesses The Naive Bayes
Sentiment Analysis Based based approach is proposed for analyzing field classifier. a test dataset's class framework will
on Extended Sentiment Chinese sentiment. The extended mood accurately and quickly. assign zero
Dictionary". IEEE Access, dictionary combines the standard mood It can be used to solve probability to any
vol. 7. pp. 43749-43762. dictionary with additional field mood words multiclass prediction identifiable variable
2019. and polysemic field mood words. The Naive problems since it in the experiment
Bayesian domain classifier is employed to works. data set and will be
classify the textual field, including polysemic unable to foresee
emotional terms, enabling the determination anything in this
of the emotion polarity associated with the regard if it is not
words. present in the

151
training dataset.
(continued)
 152
TABLE 9.1 (Continued)
Literature Survey
Algorithm
Publication Details Methodology Used Advantages Limitations
Y. Yue et al.. "Sentiment In this chapter, it deliver a approach that is used SVM. Improved the F- • Overfitting, data
Classification Based on to increase the classifier's overall accuracy Adaboosted measure of sentiment leakage,
Part-of Speech and Self when classifying Twitter data. For the same decision prediction and overall
Attention Mechanism". reason, we employ preprocessing techniques Tree and classification accuracy.
Institute of Electrical and to ensure that the training process receives decision tree.
Electronics Engineers, valid data. Our proposed method categorizes
Access, vol. 8. 2020. tweets as good or negative, which improves

loT and Machine Learning for Smart Applications

sentiment analysis and can be used to inform
future decisions.
Sentiment Analysis of Airline Tweets Using Machine Learning 153

Multinomial Naive Bayes is an alternative. The Naive Bayes procedure’s primary pur­
pose is to determine the possibilities of classes dispensed to texts using the combined
possibilities of arguments and classes. We will also employ regular expressions to
ensure our model accurately represents modern slang and emojis.

9.5 SYSTEM ANALYSIS

9.5.1 Contemporary System Limitations

Some of the present systems use the following methods for the classification of text
using sentiment: machine learning procedures similar to random forest, SVM, linear
regression, etc. In traditional systems the algorithms are repetitive, like random
forest, SVM, and linear regression. The difficulty we face while using SVM is that it
does not implement very well when the data set is additional wide-ranging, i.e., target
classes overlap.

9.5.2 Proposed System

The suggested technique uses a machine learning methodology similar to linear
regression and multinomial Naive Bayes with regular expressions together to pro­
gress the accurateness of the classical.

9.5.3 Objective
Implementing regular expression and machine learning together will help progress
the accuracy of the classical. Here, we create a list of positive and negative sentiments
using regular expressions. This list includes all the modern text abbreviations, slang,
and emojis. We will create a flag for each sentiment by looking at all patterns defined
in the list. After the above step, we get a new dataset, and after applying the ML
algorithm to the new dataset, we get better accuracy. Also, a comparative study will
be conducted to display the actual difference in the classification when a machine
learning system is used and when the machine learning system with regular expres­
sion is used.

9.5.4 Proposed Architecture

The flow of architecture is explained in Figures 9.2 and 9.3, where we begin after
taking the dataset of airline review tweets. Preprocessing and morphological feature
extraction are performed on this dataset. We try to remove the noise and unwanted data
from the dataset. Next, we will use Porter stemming for keywords in the processed
datasets. We use Inverse Frequency-Targeted Document Frequency to check the pos­
ition or relevance of string representations (words, expressions, lemmas, etc.) in a
dataset or data. Next, it will practice regular expression to make an incline of negative
and positive seeming words and emojis. In the next step, we will try to flag our
dataset tweets depending on their positive or negative attributes. Next, it will train our
154 IoT and Machine Learning for Smart Applications

FIGURE 9.2 Flow process for proposed system.

system using linear regression and multinomial Naive Bayes to categorize the text in
as positive or negative.

9.5.5 Dataset Pre-processing

To preprocess our dataset by removing unwanted data and noise. In the next step,
we will do stemming and lemmatization of the dataset. Stemming is a progression
of extracting or eradicating the last characters of a word, which frequently results
in inaccurate interpretations and spellings. Considering the context, the lemma
converts the word into its expressive base form, which is recognized as the lemma.
Here, we use Porter Stemmer and Word Net lemmatizer. We will clean our dataset
by removing abbreviations like @, #, *.%, $, &, etc. We will also use TF-IDF to
find the furthest communal word used in the article or text. TF and IDF combine the
two distinct metrics of TF and IDF. When there are numerous documents, TF and
IDF are utilized. This is constructed on the concept that uncommon words reveal
more about a document’s content than terms that are frequently utilized throughout
all publications. TF and IDF are calculated with the following formulas: where d
represents a document, ‘N’ signifies the total quantity of credentials, and df is the
quantity of credentials comprising the phrase t. TF-IDF is a word frequency statistic
highlighting the most intriguing words. The scores indicate distinct terms in a given
document in Figure 9.4.
Sentiment Analysis of Airline Tweets Using Machine Learning 155

FIGURE 9.3 Proposed architecture.

FIGURE 9.4 Stemming of words.

We will also use TF-IDF to determine the most communal word incorporated
in the file or text. TF and IDF combine the two distinct metrics to process further
stages. When there are numerous documents, TF-IDF are utilized. This is constructed
on the concept that uncommon words reveal more about a document’s content than
words that are frequently used across all publications. TF-IDF is calculated with
the following formulas: where d signifies a document, “N” is the overall quantity
of credentials, and df denotes the quantity of documents comprehending the tenure t
indicated in Equations 9.1, 9.2, and 9.3. TF-IDF is a word frequency score that aims
156 IoT and Machine Learning for Smart Applications

to highlight more fascinating terms. The scores have the effect of highlighting spe­
cific words in a given manuscript.

TF(t d) = number of times t aPPears in d

(9.1)
totalnumberof terms ind

N
IDF (t) = log-— (9.2)
1 + df

TF_IDF (t, d) = TF (t', d') * IDF (t') (9.3)

9.5.6 Classification
The final step in the process is applying the classifier used here, which is Multinomial
Naive Bayes, and linear regression with regular expression. Multinomial Naive Bayes
is a popular method for categorizing documents according to a statistical assessment
of their material, as seen in Equation 9.4. It offers a feasible alternative to “heavy” AI-
based semantic evaluation while considerably simplifying textual data classification.
The Bayesian probability p1(Ck I W):

P (Ck) * P (Ck । W)
p1(Ck |W)= (9.4)
P (W)

The fundamental tenet of the Naive Bayes is that each characteristic in W separately
influences the probability that S fits to Ck.
p1(Ck I W) Equation 9.5 can also be given as:

prior * likelihood
Posterior = (9.5)
evidence

The likelihood (W I Ck) and estimated prior (Ck) both proportionally contribute to the
outcome that Ck is the class of S.
Text search strings can be expressed using a language known as regular expressions
(REs). RE allows us to compare or identify other strings or groupings of strings by
encoding specified syntax into a pattern. Similarly, regular expressions are utilized to
search texts in MS WORD and UNIX, as shown in Figure 9.5.

9.5.7 Result Analysis

Usingregularexpressions,wewillfindthecommon100words.Wh767hg5y6yt5545565e
have written patterns that include positive emojis, positive words, etc.; here, we have
created a list of complex patterns for positive text. Here, we are making an ML model
using MultinomialNB(). Then, we have done TF-IDF vectorization of the text. We
have taken the max feature to ten because we want to train our model on the ten
Sentiment Analysis of Airline Tweets Using Machine Learning 157

# List of sample keywords

keywords = ["fLight", "canceL", "time", "heLp", "hoLd", "pLane", "caLL",
"gate”, "hour", "thank", "stiLL" , "d", "V, "0",
"0", "0", "V", "0", "&']
pattern = r'\b(?:' + join(keywords) + r')\b'
regex = re.compile(pattern, re.IGNORECASE)
# Example tweets
tweets = [
"My fLight was conceited and I need heLp.",
"I've been on hotd for over an hour, can someone pLease heLp?",
"The pLane is stiLL at the gate, not sure what the deLay is.",
"Thank you for the update on the fLight time.",
"I tried to caLL customer service but no one answered.",
]
for tweet in tweets:
matches = regex.findall(tweet)

FIGURE 9.5 Cross model 1.

import re

# List of positive emojis and words

positive_emojis = [
& , S', 'S', 'V, '4', '&'
]

positive_words = [
r '\bthanks\b', r'\bthank\s+you\b', r'\bawesome\b', r' \bgood\b', r'\bgreat\b',
r'\bexceLLent\b', r'\bamazing\b', r'\bhappy\b', r'\bfantastic\b', r'\bLove\b', r'\bjoy\b'
]

positive_pattern = .join(positive_emojis + positive_words)

positive_regex = re.compile(positive_pattern)
text = "I am so happy! S This is fantastic! Thanks a Lot!
matches = positive_regex.findall(text)
print("Positive matches found:", matches)

FIGURE 9.6 Cross model 2.

most occurring words. After training our dataset on MultinomialNB(), we print the
accuracy of the model in Figure 9.6.

9.5.8 Analysis
The results of our model outperformed previous ones significantly. A comparison
between our developed models and multinomialNB is illustrated in Figures 9.7 and
9.8. While the current state-of-the-art precision value for multinomialNB stands
at 0.82, our model demonstrates a precision value of 0.86. Furthermore, our
model exhibits a recall value of 0.52 for positive and 0.90 for negative sentiment,
representing a significant improvement compared to the state-of-the-art value of 0.22
for multinomialNB (as shown in Figures 9.9-9.12). Moreover, our model achieves an
F1 score of 0.56, which is notably higher than the state-of-the-art score of 0.35 for
multinomialNB shown in Tables 9.2-9.4.
158 IoT and Machine Learning for Smart Applications

9.5.9 Result Before Our Approach

Linear Regression

data = {
•precision': [0.81, 0.82, 0.83, 0.82, 0.82],
'recaLL': [0.99, 0.22, 0.60, 0.82, 0.90],
'fl-score': [0.89, 0.35, 0.82, 0.62, 0.80],
'support': [897, 258, 1155, 1155, 1155]
}
labels = ['Negative', 'Positive', 'accuracy', 'macro avg', 'weighted avg']

FIGURE 9.7 Positive and negative sentiments with various parameter for linear regression.

Classification Report Heatmap

Precision Recall fl-score

Metrics

FIGURE 9.8 Positive and negative sentiments with accuracy for linear regression.

data = {
'precision': [0.90, 0.60, 0.86, 0.75, 0.85],
'recall': [0.93, 0.52, 0.86, 0.72, 0.86],
'fl-score': [0.91, 0.56, 0.86, 0.74, 0.85],
'support': [897, 258, 1155, 1155, 1155]
}
labels = ['Negative', 'Positive', 'accuracy', 'macro avg', 'weighted avg']

FIGURE 9.9 Positive and negative sentiments with various parameter for MultinomialNB.
Sentiment Analysis of Airline Tweets Using Machine Learning 159

Recall
Metrics

FIGURE 9.10 Positive and negative sentiments with accuracy for MultinomialNB.

9.5.10 Result After Implementing Proposed Approach

precision recall fl-score support

negative 0.91 0.91 0.91 952

positive 0.59 0.60 0.59 203

accuracy 0.86 1155

macro avg 0.75 0.75 0.75 1155
weighted avg 0.86 0.86 0.86 1155

Confusion Matrix: [[867 85]

[ 82 121]]

FIGURE 9.11 Positive and negative sentiments with various parameter for linear regression.

TN+ TP
Accuracy = (9.6)
TN+FP+TP+ FN

TP
Precision (P) = (9.7)
TP + FP

TP
Recall (R) =-------- (9.8)
TP+ FN
160 IoT and Machine Learning for Smart Applications

-0 65

-0 60
Recall
Metrics

FIGURE 9.12 Positive and negative sentiments with accuracy for linear regression.

TABLE 9.2
Twitter Airlines Review

Accuracy
Dataset Algorithm Used Initial Accuracy (%) Final Accuracy (%)
Twitter Airlines review Linear regression 82.0 86.3
Twitter Airlines review MultinomialNB 82.3 86.9

TABLE 9.3
For Positive Sentiment

Precision Recall F1_score Accuracy

Linear regression 59 60 59 86
MultinomialNB 60 52 86 86.9

TABLE 9.4
For Negative Sentiment

Precision Recall F1_score Accuracy

Linear regression 90 93 91 86
MultinomialNB 91 91 91 86.9
Sentiment Analysis of Airline Tweets Using Machine Learning 161

Precision (P) * Recall (R)

F1 - score = (9.9)
Precision (P)+ Recall (R)

Accuracy
The distinctiveness lies in the ratio of the combination of true negative (TN) with
true positive (TP) divided by the sum of true positive (TP), false negative (FN), false
positive (FP), and true negative (TN). The accuracy of the projected classification is
achieved and detailed in Tables 9.2-9.4 and Equation 9.6.

Precision
It is distinct as the proportion of the quantity of tweets related to the quantity of
inappropriate data retrieved. The performance of the projected classification
accomplishes high when the rate of precision is high, and it is low when the precision
rate is low, as depicted in Tables 9.2-9.4 and Equation 9.7.

Recall
It is distinct as the proportion of (TP)True positive by the sum of true negative (TN)
and false negative (FN). The recall of the projected classification is accomplished and
mentioned in Tables 9.2-9.4 and Equation 9.8.

F1 Score
It is distinct as the ratio of meticulousness multiplied by recall(R) and the sum of
precision with recall. The Fl score of the projected classification is accomplished and
mentioned in Tables 9.2-9.4 and Equation 9.9.

9.6 CONCLUSION
In this chapter, sentiment scrutiny of airline tweets consuming machine learning
systems and regular expression has been training itself. Our model examines a total
of 11541 tweets from 6 distinct airlines and we implement a technique of multi­
nomial NB with regular expression. Also, the proposed concept in the chapter, which
has much better accuracy as a result, is more efficient and has improved accuracy,
which was successful. The retrieved characteristics are trained. And classified Twitter
airlines review, got output as for initial accuracy for linear regression is 82% and final
accuracy is 86.3%, for MultinomialNB, the output as for initial accuracy for linear
regression is 82.3% and final accuracy is 86.9%. The result analysis of positive senti­
ment using linear regression, we achieved 59% for precision, 60% for recall, 59% for
F1 score, and a final accuracy is 86%, and similarly in the result analysis of positive
sentiment using MultinomialNB we achieved 60% for precision, 52% for recall, 86%
for F1 score and accuracy is 86.9%. For the result analysis of negative sentiment using
linear regression, we achieved 90% for precision, 93% for recall, 91% for F1 score,
and a final accuracy is 86%, and similarly, result analysis of positive sentiment using
MultinomialNB we achieved 91% for precision, 91% for recall, 91% for F1 score,
and accuracy is 86.9%.
162 IoT and Machine Learning for Smart Applications

REFERENCES
[1] James Schaffer and Brandon Huynh, “An Analysis of student Behavior in Two
Massive open online courses,” IEEE/ACM International Conference on Advances in
Social Network analysis and mining (ASONAM), 2016.
[2] Kevin Andrey Ferreira and Liedy’s del Carmen, “Analysis of Behavior of Customers
in the Social Networks using Data Mining Techniques,” IEEE/ACM International
Conference on Advances in Social Network analysis and mining (ASONAM), 2016.
[3] C. Verma, V. Stoffova, Z. Illes, S. Tanwar and N. Kumar, “Machine Learning­
Based Student’s Native Place Identification for Real Time,” IEEE Access, vol. 8,
pp. 130840-130854, 2020, DOI: 10.1109/ACCESS.2020.3008830.
[4] M. J. Adarsh and P. Ravikumar, “An Effective Method of Predicting the Polarity
of Airline Tweets Using Sentimental Analysis,” 2018 4th International Conference
on Electrical Energy Systems (ICEES), 2018, pp. 676-679, DOI: 10.1109/
ICEES.2018.8443195.
[5] F. Es-Sabery, A. Hair, J. Qadir, B. Sainz-DeAbajo, B. Gartfa-Zapirain and I.
D. L. Torre-D^ez, “Sentence-Level Classification Using Parallel Fuzzy Deep
Learning Classifier,” IEEE Access, vol. 9, pp. 17943-17985, 2021, DOI: 10.1109/
ACCESS.2021.3053917.
[6] A. Ishaq, S. Asghar and S. A. Gillani, “Aspect-Based Sentiment Analysis Using a
Hybridized Approach Based on CNN and GA,” IEEE Access, vol. 8, pp. 135499­
135512, 2020, DOI: 10.1109/ACCESS.2020.3011802.
[7] X. Lei, X. Qian and G. Zhao, “Rating Prediction Based on Social Sentiment From
Textual Reviews,” IEEE Transactions on Multimedia, vol. 18, no. 9, pp. 1910-1921,
Sept. 2016, DOI: 10.1109/TMM.2016.2575738.
[8] Z. Chen, F. Lu, X. Yuan and F. Zhong, “TCMHG: Topic-Based Cross-Modal
Hypergraph Learning for Online Service Recommendations,” IEEE Access, vol. 6,
pp. 24856-24865, 2018, DOI: 10.1109/ACCESS.2017.2782668.
[9] HT. Manshu and W. Bing, “Adding Prior Knowledge in Hierarchical Attention
Neural Network for Cross Domain Sentiment Classification,” IEEE Access, vol. 7,
pp. 32578-32588, 2019, DOI: 10.1109/ACCESS.2019.2901929.
[10] K. Abdalgader and A. A. Shibli, “Experimental Results on Customer Reviews
Using Lexicon-Based Word Polarity Identification Method,” IEEE Access, vol. 8,
pp. 179955-179969, 2020, DOI: 10.1109/ACCESS.2020.3028260.
[11] H. S. Lee, “Exploring the Initial Impact of COVID-19 Sentiment on US Stock
Market Using Big Data,” Sustainability, vol. 12, no. 16, p. 6648, 2020, DOI: 10.3390/
su12166648.
[12] G. Xu, Z. Yu, H. Yao, F. Li, Y. Meng and X. Wu, “Chinese Text Sentiment Analysis
Based on Extended Sentiment Dictionary,” IEEE Access, vol. 7, pp. 43749-43762,
2019, DOI: 10.1109/ACCESS.2019.2907772.
[13] M. Rathi, A. Malik, D. Varshney, R. Sharma and S. Mendiratta, “Sentiment Analysis
of Tweets Using Machine Learning Approach,” 2018 Eleventh International
Conference on Contemporary Computing (IC3), pp. 1-3, 2018, DOI: 10.1109/
IC3.2018.8530517.
[14] W. Tai, T. Zhong, Y. Mo and F. Zhou, “Learning Sentimental and Financial Signals
With Normalizing Flows for Stock Movement Prediction,” IEEE Signal Processing
Letters, vol. 29, pp. 414-418, 2022, DOI: 10.1109/LSP.2021.3135793.
[15] K. Cheng, Y. Yue and Z. Song, “Sentiment Classification Based on Part-of-Speech
and Self Attention Mechanism,” IEEE Access, vol. 8, pp. 16387-16396, 2020,
DOI: 10.1109/ACCESS.2020.2967103.
 Smart Workspace
10 Automation
Harnessing IoT and AI
for Sustainable Urban
Development and Improved
Quality of Life
Sapna R., Preethi, Pavithra N., Manasa C.M.,
and Raghavendra M. Devadas

10.1 INTRODUCTION
Several individuals spend a significant amount of time in workplaces these times.
Everybody should feel at ease in the office since the working environment seriously
influences how successfully employees accomplish their tasks. So, convenience is
essential and required in the workplace. When technology was at its greatest in the
past, it meant having an electronic pager and a landline connection; today, it means
having an electronic tablet linked to the web. A smart office is one that empowers
life with ease and comfort for its people and clients, empowering them yet enhan­
cing their capability to keep themselves connected to one another. Organizations are
harnessing the power of modern technology and implementing effective strategies to
drive staff efficiency and output. As the physical barriers are being breached, a multi­
faceted and competitive atmosphere prioritizing innovation and creativity is forming.
Intelligent growth zones are rapidly emerging worldwide, so smart offices quickly
become necessary.
A “smart workplace” ensures IT resources and infrastructural facilities are used
as effectively and efficiently as possible. In other terms, offices are automated in
the current generation of information technology. An accessible, innovative technical
setting is required. Thus, office automation replaces processes that are more access­
ible and encourages open information exchange, which creates the opportunity to
have a substantial effect on how industries as well as enterprises operate. The system’s
usage of various communication tools and efficient automated systems demonstrates
the favorable effects on a company’s or lender’s long-term growth. The benefit of a
smart office is eliminating internal reporting procedures, such as employee arrival and
departure times, through an open office layout. Improving cooperation and communi­
cation can raise efficiency, which influences the outcome.

DOI: 10.1201/9781032623276-10 163

164 IoT and Machine Learning for Smart Applications

A smart workplace needs to be developed to maximize employee and worker

potential. This is not a supernatural event; it is just revolutionary innovation and cre­
ative thinking that better meets people’s requirements. Direct conversation and simple
recordkeeping are among the benefits of office automation. One of the key drivers
of worldwide energy consumption is the building sector. It uses almost a third of all
the energy consumed. Like a smart house, a smart building uses sophisticated sensor
data collection and analysis to maximize comfort, efficiency, and safety. Complicated
machine devices, sophisticated control systems, and other elements are found in
modern buildings to increase occupant safety and efficiency.
A fantastic system of linked subsystems might be considered a smart building.
All a building’s systems and appliances must be connected to be considered a
smart building. The managers can utilize it to view information and quickly and
accurately establish judgments. Smart buildings can be made using a variety of
methods. The automated centralized control of an office infrastructure’s various
components, such as light, heating, ventilator, cooling systems, and other systems,
is accomplished via an automatic building management system. The objectives
of smart building systems are comfortable structural framing functioning, lower
energy consumption and operating costs, and extended utility developmental
stages.
By controlling the power consumption in business buildings, considerable energy
can be saved. These problems lead to the concept of the “smart office.” This smart
office design concept can be used for the entire structure to use less energy. In these
days, security is essential. Due to the introduction of rapidly developing technology,
users now need strong security solutions. The systems need to be very secure. These
identifying techniques include user password-based systems, ATMs, other smart
cards, and many others. Nevertheless, hacking assaults, theft, and credential forget­
ting render these systems insecure.
Radio-frequency identification (RFID) has also been employed for security
reasons. However, RFID has the drawback that it can be hijacked, or another indi­
vidual can enter using that ID. This is definitely not safe. Despite all these flaws and
errors, the most effective and dependable security solution is a biometric or biometric
authentication-based identity.
That would save energy and improve employee satisfaction. A complete smart
office system with targeted lighting, heating, monitoring, and alerts is constructed
using this approach. Several strategies are used in smart energy solutions to create
energy savings that benefit from organizational conduct. This strategy, which has
been implemented, exemplifies how to cut back on energy use by realizing the sig­
nificant influence that people have on it.
Due to the extensive use of innovative technologies in professional settings,
smart office concepts have gained popularity. Even though companies have been
implementing smart office ideas to give their customers practical and efficient
workspaces [1], the studies have given less attention to the user’s perspective. It is
well known that relevant issues and aspirations should be considered when developing
workplace settings [2]. Still, it is unknown what intentions and interests people have
for automation and control ideas, as well as what specific features justify smart office
Smart Workspace Automation 165

concepts and differentiating themselves from other non-smart workplace types. To

comprehend the user requirements for smart office spaces, empirical information
would be of great social and scientific relevance.

10.2 STATE OFTHEART

This section examines how the Internet of Things might boost production, worker
productivity, and safety while enhancing performance evaluation procedures. It
looks at how Internet of Things (IoT) can deepen the relationship between manage­
ment and employees. System of work at manufacturing facilities. After reading the
relevant literature, IoT aids in observing and controlling worker behavior, enhan­
cing performance, and granting workers more independence. Analyzing IoT’s role
in worker safety and productivity is the first step in understanding how it may
benefit workers. Organizations may not even be aware of all the safety challenges
they face. IoT aids with the use of precise sensors. These problems are identi­
fied, advance warnings are given so that action can be taken, and rescue efforts
are facilitated. Tracking workers’ vitals and making suggestions in line with them
not only helps to improve safety standards but also worker productivity. Building
this dynamic workspace would be practically impossible without the IoT, so we
explore how it works. We also examine alternative IoT-based performance evalu­
ation methods. Conventional techniques of measuring employee performance are
occasionally prone to manager biases and are only undertaken occasionally. IoT,
on the other hand, provides constant feedback and is impartial. Conventional HR
practices will not be fully abandoned; rather, they will be strengthened with the
help of consistent, dependable feedback.
We can answer the crucial queries of how IoT can assist and how it is used in
human practices by going over these two subjects. This also establishes the frame­
work, for instance, in which IoT is employed to track employees’ geolocations
when required to precisely understand how much time each worker contributes
to work. Reinforce them with the aid of ongoing, accurate feedback. The 4th
Industrial Revolution is the present phase of technological progress characterized
by an emphasis on AI-driven technologies. Data-driven AI technologies are cur­
rently fostering an environment of Industry 4.0, which follows the 3rd Industrial
Revolution launched by the Internet and mobile Internet. Industry 4.0 describes
the current technology trend to automate processes and interchange information.
Numerous inexpensive tools can gather real-time data and send it through the IoT.
As a result, using a remote control is possible.
The IoT makes it possible for machines and devices to communicate with one
another and generate huge quantities of data, potentially revealing significant infor­
mation in a few service sectors. We can create integrated semantic systems that
enable semantic interoperability by fusing the IoT context with semantic technolo­
gies. The development of research systems combines embedded IoT, which is used to
remotely operate home appliances, particularly electronic equipment, with embedded
ESP8266, which acts as a control with real-time firebase. ATMEGA16 microcon­
troller and 20
166 IoT and Machine Learning for Smart Applications

18 Fourth International Conference on Computing, Engineering, and Design [3].

The local and global Message Queuing Telemetry Transport (MQTT) servers can be
used to construct a remote temperature monitoring system that employs the MQTT
protocol and a database that serves as a mobile Backend Service to store data [4].
MQTT server is used, anyone with an internet connection can access the tempera­
ture from anywhere at any time. The IoT technology can be utilized to protect homes
from unforeseen issues, including fire, theft, temperature conditions, and motion
gestures [5].
One of the pioneering studies to emphasize the significance of the sustainable
development of IoT and how innovative technology should not be a goal in and
of itself is that of the authors in [6]. Hoang et al. [7] have evaluated how various
Information and Communication Technology (ICT) solutions could be leveraged to
lower energy usage. The studies by Adhikari et al. [8] have more recently discussed
incorporating sustainable energy sources into smart developments and any potential
difficulties. As an expanding dissertation topic on the need aspect of a smart grid,
the hurdles to the deployment of electric vehicles in relation to energy demand have
been exhaustively analyzed by the researchers in Sanguesa et al. [9]. Analogously,
the examinations conducted by Omitaomu et al. [10] assessed the matter, albeit in the
specific domain of smart cities. The studies by Yigitcanlar et al. [11] have offered a
thorough assessment on the subject, and numerous AI applications have been created
for the smart grid. Still, it has not yet been determined if the adoption of AI can
address environmental issues. The researchers in Crossley and Beviz [12] analyzed
the general limitations of AI in smart buildings.

10.3 USING AUTOMATION TO ADDRESS ENERGY ISSUES

Electricity production, public transit, and construction will all be considered, as
they are the main factors that determine emissions of greenhouse gases. Figure 10.1
illustrates the key technology components that deal with energy in Smart Office.

FIGURE 10.1 Smart office.

Smart Workspace Automation 167

10.3.1 Smart Grid

A smart grid, a power arrangement that relies upon sensors, communications,
controllers, and information systems, uses IoT ideas to harmonize the power grid’s
current activities and make them more efficient and dynamic in their administration.
The IoT, specifically wireless sensor networks, tracks solar energy production. In
the agricultural sector, IoT technology has advanced to facilitate the identification of
viable agricultural land and the selection of suitable crops by utilizing Raspberry Pi.
Within the healthcare domain, IoT is employed to monitor cardiac activity precisely,
specifically focusing on heart rate.
The idea of a smart grid is inextricably linked to the emergence of rapidly chan­
ging sustainable assets with transferring belief toward alternatives to fossil fuels, all
while keeping supply and demand manageable [13]. A bi-directional power flow is a
common cornerstone for recognition in a smart grid, which implies that customers are
also energy manufacturers, abandoning the standard function of inert users [14]. The
main challenge for the smart grid is the networks themselves, as the majority were
built to take in massive fossil fuel-powered facilities, supplying power to customers
only in one way [15]. Figure 10.2 shows a general view of a typical power grid.
Three vital elements required for this bidirectional flow are demand-side manage­
ment, storage technology, as well as real-time infrastructure management. All these
elements promote technical assistance. Four types, such as load prediction, power
grid stability evaluation, defect analysis, and security concerns, have seen the most
use of AI [11].
As evidence for the latter group, the incorporation of information technologies
into the electric grid itself could create security issues because they can facilitate
hacks meant to interfere with regular, daily operations. Artificial intelligence (AI)
techniques were developed to recognize when such assaults occur. The develop­
ment of a smart grid can be seen in the context of a smart workplace as a basic
building block for addressing the environmental concerns and resource scarcity that
are becoming increasingly widespread globally.
Variable pricing is one method used to match energy supply to demand. For
example, a rise or drop is synchronized due to a rise or fall in requests made for
the energy. As a result, during peak times, the cost is advanced even though energy
plants are operating at near-maximum ability, whereas during normal hours, the cost
is less. This means clients are urged to utilize equipment when energy demand is low.
It’s true that demand response efforts are designed to keep customers by providing a
monetary reward in the form of lower electricity bills. These efforts’ versatility is the
capable of accommodating the variations of renewable energies and enabling their

FIGURE 10.2 A general view of a typical power grid.

168 IoT and Machine Learning for Smart Applications

greater depth [16]. AI-powered solutions have been created to help such practices
in this scheme. Boza and Evgeniou [17] suggested an RL algorithm to stabilize ser­
vice suppliers’ profits and clients’ savings in order to attain power system reliability.
Data centers are a potential load management sector where AI can be used [18]. The
enormous volume of data a smart office should produce is among its most crucial
features [19]. Nevertheless, the establishment and growth of such a smart workplace
may face environmental challenges related to the collection and storage of data. Data
centers are energy-intensive businesses that consumed 1% or more of the world’s
electricity in 2018. Their efficiency is based on the consolidation of multiple servers.
According to the research by Crawford [20], data centers will use 2.13% of the total
electrical supply by 2030 in 2021. Mytton [21] reports a warning against statements
made by businesses that they are carbon neutral because this may be the result of
buying carbon credits instead of using renewable sources of energy.
From an environmental standpoint, the sustainable development of data centers is
inextricably linked to the use of renewable resources to influence their operations. It is
crucial to note that exorbitant water usage is a related problem because water is used
to keep down data centers, which can have a significant role in energy utilization. Yet,
it is also employed in oblique ways to generate power [22]. This is significant to the
water shortages due to climatic change, as it has been disclosed that the global water
supply, such as snow and ice, has reduced by 1 cm per annum over the last 20 years [23].
Although service providers and lawmakers have taken steps to reduce these environ­
mental issues [22], optimizing energy usage can help to solve this challenging problem.
Li et al. [23] recommended an algorithm to optimize the data center. Cooling systems
and reducing associated costs. A related real-world AI application in data centers is the
ML method proposed by DeepMind for Google, which reduced the expanse of energy
needed for the cooling system by 40%. According to reports, Google and DeepMind will
work together to make the technology accessible to commercial organizations [24,25].
Energy storage is another method for achieving both supply and demand stability.
There are numerous types of energy storage systems, including mechanical (like
flywheels), electrochemical, thermal, and chemical. They may also be split into two
groups based on where they apply. Grid-scale implementations appropriate for mass
storage are accessible. Demand-side applications, on the other hand, are confined to
small regions. Demand-side storage is frequently used with on-site energy produc­
tion, as is typically with some sustainable energy sources, like wind and solar. A wide
range of storage technology applications for AI can be thought of. AI has been used,
for example, to create and evaluate the storage systems themselves, such as when
choosing the requirements for lithium-ion batteries to increase their individualism.
An algorithm was created to evaluate the suitable location and measurements of an
energy storage system in the context of improving management in a business signifi­
cantly an electric grid using renewable energy sources, as doing so should improve
reliability and minimize power shortfalls [26].

10.3.2 Smart Monitoring of Employees

The stimulation of various electronic monitoring structures today makes it possible
to improve management in a business significantly. Various firms are increasingly
Smart Workspace Automation 169

adopting technology to observe employees and improve staff effectiveness elec­

tronically. This is correlated to the enhancement of any business. This study aims
to examine how several employee performance indicators are affected by electronic
surveillance. Three different organizations’ present surveillance systems have been
noticed, and for the research, 20 staff members from each company were chosen as a
sample. It has been found that these systems heavily rely on the employees’ logging
in and out times and that their punctuality and sincerity significantly affect both their
individual and collective performance. To increase the productivity rate of the com­
pany and employees, no surveillance monitoring has ever been established that can
track the complete activities of the staff members on the resources in office time. To
effectively monitor any computer system during working hours, the solution provided
by Chinyere and Chiemela [27] analyzes the required observational parameters and
incorporates a keylogger into the system.
There are two distinct types of keyloggers: hardware-based and software-based
keyloggers. Software keyloggers can be installed on a computer either locally or
remotely. While hardware-based physical keyloggers are simple to find whether a
user finds what devices/resources are linked to the computer, software keyloggers
are invisible to the human eye. Software-based keyloggers can reproduce using
a virtual system, hypervisor-based or virtual system manager, operate as a key­
board driver, or use other methods to access the operating system of the targeted
machine [27].
Bernstr0m and Svare [28] examined the covert measures employers take to
monitor their workers’ behavior. E-monitoring and surveillance are the main
approaches employed. Although it has been noted that the system in use is paradox­
ical, employers’ goals are to accomplish their specific objectives and preserve the
system’s performance, effectiveness, and data protection. Practically, it becomes a
disruption to their employees’ privacy. According to research, more organizations
are implementing this type of surveillance since the effects of this realization lead to
expansion. The author further concluded that employers should try to uphold appro­
priate corporate conduct to ensure productive management.
Jeske and Kapasi [29] examine workplace surveillance efficacy and strive to
distinguish between command and monitoring, as well as between fundamental
inspiration and proficiency. Additionally, structural equation modeling was used
in this study to examine the anticipated Norwegian employee sample samples.
The method employers use to introduce and administer monitoring to their work­
force will determine if it has a positive or negative effect on them. According to
the findings, workers feel less convicted when they know they are being watched.
However, the study reported in the chapter shows that employees are still less
motivated and have less faith in the system, even once a surveillance system is
in place. A system is necessary for employees to better understand how security
measures are implemented.
Ahmad et al. [30] examined the information found in the electronic performance
monitoring method. The primary focus of this study is on how data are generally used
following the use of monitoring techniques. Surveillance to increase security and
monitor employee performance may help the organization expand. Still, it has also
been shown to occasionally breach employees’ privacy and hurt how they view the
170 IoT and Machine Learning for Smart Applications

company. This indicates that the distinction between the methods employed to assess
the outcome is an important consideration for management edification and the formu­
lation of moral judgment.
Vujovic and Maksimovic [31] suggested authoring a report to elaborate on the
influence of performance monitoring, information/data security, and additional social
immersion coaching elements on employees’ behavior regarding safety assertion.
The safety pledge represents the actions taken to protect employees’ diligent work
and information arrangements. The behavior is very desired because it relates to a
personal aspect within the context of information safety. The authors concluded that
implementing information security monitoring can enhance the cultural behavior
of safety assertion. The study concluded that when employees’ behavior is deemed
inconvenient, safety procedures must be abandoned.
Many earlier studies on monitoring focused on the benefits and drawbacks of
e-monitoring implementation and employees’ reactions. Many businesses have
introduced e-surveillance without the knowledge of their employees, which has a nega­
tive effect and makes them worry about their privacy. Most systems use monitoring
systems to keep track of employees’ time records. By utilizing the biometric tech­
nology built within their system, practically every firm heavily relies on this strategy.
As a result, a system that can monitor employees’ computer use while they are at
work is needed, in addition to being able to record their arrival and departure times.

10.3.3 Smart Office Building with Controlling Lights, Fans,

and Other Electronic

A Building automation system lowers energy usage and makes utilizing various
appliances more convenient. Thanks to the energy-saving strategy, building auto­
mation today makes a living quite simple. All electrical and technological home
appliances and devices will be controlled remotely through a wireless connection.
Individual control device systems, distributed control device systems, and centrally
controlled device systems are all possible configurations for automation systems
[32]. The simplest home automation devices are individual control devices, which
are programmable gadgets that may be configured to user preferences. On the other
hand, distributed and centrally controlled systems, which build automation systems,
have remote communication and device control capabilities. The distinction between
distributed and central systems is that the latter has a controller to centrally control
all the devices, while the former does not. The disadvantage of a centralized control
system is the breakdown when the controller breaks down [32].
Commercial building automation systems are pricey because they require
specialized hardware, components, and installation. Utilizing open-source platforms
and IoT sensors to perform specialized automated home systems that is tailored to the
demands of the end user is an alternative strategy. The cost of IP-based devices and
customer inexperience with open-source home automation systems are the two main
drawbacks of this strategy. This automation needs to install multiple such pricey IP­
based devices, which are out of the price range of low-income homes.
The image shows that the system’s primary implementation components are
sensors, regulating mechanisms, and actuators. The sensors’ data regarding light,
Smart Workspace Automation 171

motion, temperature, and other sensory features is transferred to the primary controlling
devices. Various sensors, such as photodetectors, level sensors, pressure transducers,
transformers, infrared sensors, and temperature sensors or thermistors, require add­
itional signal processing apparatus to communicate with the main controller.
Controllers are designed to be coupled to control devices, such as program­
mable logic controllers, which get data from sensing devices and perform operations
upon them in accordance with a program. This software could be altered by the
load operations. Different analog or digital inputs and output units, such as sensing
devices, actuating devices, and other devices can be connected to the programmable
controller.
The last devices to regulate the appliances are called actuators, which com­
prise limit switches, relays, motors, and other controlling mechanisms. In the home
automation system, connectivity is essential for remote access to these operations.
Additionally, this smart home system provides scheduling, power features, and con­
stant surveillance via video surveillance with cameras. This is the best treatment—
even for the old. A simple office building automation system is shown in Figure 10.3.
Typically, it takes place in an automated home setting. It can be divided into sev­
eral sorts. A smartphone can be used to control some of them remotely. Second, it can
command a motor or actuator that assures security, like a door lock. Nowadays, many
building automation systems are mostly Android-based, allowing a single smartphone
to manage all household appliances. Arduino, relay modules, Wi-Fi (IEE802.11 b/g/
n)/4G, 5G, or Bluetooth modules can be used with a PC or microcontroller to quickly
create a home automation prototype.
Relay channel count is determined to quickly create a home automation prototype
based on how many appliances you want to connect to the automation system. An
Arduino board can accept up to 12 V of power. It might also be powered by a 9-V DC
battery. Arduino’s 5 V/3.3 V output pin can power Bluetooth and the relay module.
The most common asynchronous transfer method utilized by the Bluetooth
module is the universal asynchronous receiver/transmitter (UART) mode with a band
rate of 9,600-115,200. Use this mode to communicate with another device, such as
a laptop or smartphone. In this instance, a native Android app was utilized to link the
Bluetooth module with the Android phone and control the appliances.

FIGURE 10.3 An office building automation system.

172 IoT and Machine Learning for Smart Applications

The relay primarily manages electrical devices, including fans, lights, refrigerators,
and air conditioners. It offers total isolation between the high-voltage line and Arduino
or another module. Its primary coil is agitated by delivering 5 V to the coil, which is
mostly applied from the Arduino board.

10.3.4 Smart Vehicles with Automated Parking

The automated parking system combines software, a mobile application, and trans­
portation instruments for the best possible user experience. This system is made up
of a range of electromechanical devices and robots that are controlled by computers.
Without the aid of a person, vehicles are stored, retrieved from, and placed into open
multi-depth parking spaces using conveyance equipment, which can move both
horizontally as well as vertically. The system under consideration is characterized
by its electromechanical transport apparatus, designed within a closed structure that
may consist of a single level or multiple levels. It facilitates movements across mul­
tiple dimensions, directions, and depths, indicating a versatile and dynamic range of
transport capabilities. The system incorporates designated rooms for entry and exit,
emphasizing its functional and organized design to facilitate seamless transitions
and accessibility. The combination of electromechanical transport, closed structure,
and multi-dimensional movements suggests a sophisticated system tailored for effi­
cient and controlled transportation within various environments.
The automated parking garage has no humans driving or walking through it. High-
density parking and effective land use are made possible by this. The system effect­
ively uses the building by not requiring ramps, turning radii, or pedestrian walkways,
freeing up valuable space for additional units or amenities.
The automatic parking system functions in contrast to a standard one. It is almost
like a valetless valet parking service; it is speedier and requires less driver involve­
ment. The automated parking system has several bay rooms for parking and riding away
vehicles. The car’s driver approaches the harbor and ceases at the arrowed sign. When
a rolling door gives way, the driver can take the vehicle into the bay, where the sensing
device measures the vehicle’s measurements and then instructs the driver on how to
position the car properly on a screen in front of them. When the car has reached the
right position and is prepared for parking, the driver sees a notification on the screen.
After securing the car with the key and exiting the vehicle, the driver drives to
finish the parking process. The parking operation is started with a smartphone app, a
card swipe at the payment terminal, or a pay ticket. When the app verifies the parking
request, the bay door closes, and sensors scan the area to verify that no movement is
observed on the exterior of the vehicle. A shuttle system removes the vehicle from the
bay and stores it in the right area based on dimensions. The Bay room is now ready
for another vehicle to be parked or retrieved from safekeeping.

10.3.5 Industry Revolution

The continued Industrial changes and growth of Industry 4.0 from the start of the
Industrial Revolution 1.0 (1760-1840) have caused a change in basic assumptions
in healthcare centers provided globally. Industry 4.0 combined developments in
Smart Workspace Automation 173

computing techniques with the emergence of cyber systems, the IoT, cloud computing,
cognitive computing systems, and artificial intelligence (AI) technical platforms.
Since its introduction in 2011, innovations in Industry 4.0 have revolutionized caring
for patients at every stage, including integrating medical facilities, accelerating the
growth of groundbreaking treatments, and prompt detection.
Unfortunately, Industry 4.0 did not meet the increasing demand for “Personalized
care” despite recent automation and high efficiency satisfying the requirement for
production in large quantities with reduced human effort using intelligent produc­
tion methods. Industry 5.0, a concept first suggested in 2015, has made it possible
to personalize products and integrate more human intelligence into manufacturing.
Enhancing the interplay between smart machine systems, human intelligence, and
robotic technology is anticipated to design and deliver individualized care and
responsibility for patients with orthopedic disorders. This helps in building on the
sophistication of Industry 4.0 technologies.
Industry 5.0, unlike its predecessor Industry 4.0, shifts its focus from solely
maximizing production efficiency through automation to prioritizing the human
element in manufacturing. It emphasizes the development and implementation of
cutting-edge technologies, such as the IoT and big data analytics, to enhance human
capabilities and improve job satisfaction. While Industry 4.0 focused on replacing
human labor with robots and intelligent systems, Industry 5.0 aims to create collab­
orative environments where humans and machines work together synergistically,
leveraging each other’s strengths. With proof of recent research and technological
development, industry 5.0 will advance marketably. Over the last five years, the
manufacturing sector has witnessed significant advancements and transformations
driven by key technologies. Artificial intelligence (AI) and machine learning (ML)
have played pivotal roles in optimizing production processes, enhancing predictive
maintenance, and improving overall efficiency. Collaborative Robots (Cobots)
have been integrated into manufacturing workflows, fostering human-robot col­
laboration to streamline tasks and increase productivity. The Internet of Everything
(IoET) has interconnected devices and systems, creating a smart and interconnected
manufacturing environment. Blockchain technology has been employed to enhance
supply chain transparency, traceability, and security. Additionally, adopting digital
twins has become prominent, enabling virtual replicas of physical assets to enhance
monitoring, analysis, and decision-making in the manufacturing sector. Collectively,
these technologies signify a paradigm shift toward more intelligent, connected,
and data-driven manufacturing processes. Their use cases will change, and new
implementation areas will emerge that are essential to defining Industry 5.0. AI has
been integrated into every stage of the manufacturing value chain. AI-enhanced
human machines are a crucial development that will reestablish AI’s standing in
Industry 5.0. The level of interaction between humans and robots has increased
from simple to complex, encompassing the use of data to verify, manufacture, and
operate autonomous vehicles and the development of sophisticated storage facility­
picking robots capable of classifying commodities based on the order in which they
are obtained
With the confluence of IoT, big data analytics, and AI, the Smart Factory was born.
Industry 4.0 marked the beginning of the IT-OT combo, which is still expanding to an
174 IoT and Machine Learning for Smart Applications

extent to become IoET. IoET, taken as a whole, refers to various intricately coupled
devices.
In the next decades, these developing technologies are anticipated to grow in a
number of important fields, including:

1. AI-ML: Increased Intelligent Automation will boost office and shop floor
productivity. Businesses place a high priority on quality management, and
AI will help them do that by facilitating decisions in less time. For instance,
process manufacturer Koch collaborates with Independent Software Vendor
(ISVs) headed by AI, such as C3.AI, to deploy AI throughout the organiza­
tion. This will enable managing energy use, planning and scheduling manu­
facturing, and optimizing inventory.
2. Cobots: The employment of cobots in production and warehouse management
will boost factory productivity, leading to improved uniformity and precision
in product production. ABB, the first company to offer cobots, is leading
technological breakthroughs with its 46 years of cobot-building experience.
Cobots are being introduced to replace manual labor on the shop floor for
Automated Guided Vehicles (AGV) mounting, assembly line automation,
simple packaging, and industrial testing.
3. IoET: IoT is gradually evolving into IoET, and connected industrial
equipment is assisting producers in reducing costs, increasing asset product­
ivity, lowering downtimes, and establishing realistic supply chain visibility.
A notable example is Koch, which is utilizing invisible, IoT-enabled tapes
from US-based ISV to make its factories smarter. This clever wireless tape
transmits data like location, position, and temperature while monitoring valu­
able cargo and equipment.
4. Blockchain: Assuring operational transparency and enhancing supply chain
visibility, asset monitoring, and operational enhancements for manufacturers
are all aspects of operations in which blockchain technology is gaining
prominence. This development is not just limited to the Banking, Financial
Services, and Insurance (BFSI) sectors.
With Honeywell, this development has already taken a step forward. To
replace manual processes and paper use, Honeywell Aerospace is assiduously
investing in Blockchain to list products for sale, track stakeholder interactions,
and keep track of transactions.

10.4 CONCLUSION
The existing smart workplace principles are expanded upon in this chapter from
the perspective of users. Because cutting-edge technology and design might create
an expectation of modern workplaces with smart office ideas, the requirement for
ongoing improvement of the workplaces must be considered from a management per­
spective. This study recommends that a precise statement of the intelligent idea should
be supplied to the consumers, referring to the specific qualities, as various users may
have varied viewpoints and aspirations for smart ideas. With the smart workspace,
Smart Workspace Automation 175

employees can manage the equipment throughout the office to cut down on electri­
city waste. For the upcoming development, it is anticipated that a better system will
be developed and created with the addition of more and different outcomes to allow
for greater enhancement in attempting to control other equipment present throughout
the workplace. It can also be created with a data warehouse within every individual’s
smart working space.
All new construction will soon include sensing and communication systems to
give occupants the desired convenience. As a result, it will impact how buildings
are designed in the future, possibly even turning non-residential facilities like hotels
and airports into intelligent structures. There will be a multitude of cutting-edge
life cycle models for smart buildings. To give users of the smart building a richer
user experience, all units should be flexible enough to accommodate technological
improvements.

REFERENCES
[1] Bodker, S. “Rethinking technology on the boundaries of life and work,” Personal
and Ubiquitous Computing, vol. 20, no. 4, 2016, pp. 533-544, DOI: 10.1007/
s00779-016-0933-9.
[2] Kim, J., De Dear, R. “Workspace satisfaction: The privacy-communication trade-off in
open-plan offices,” Journal of Environmental Psychology, vol. 36, 2013, pp. 18-26.
[3] Alipio, M.I. Development of Smart Indoor Workplace System Using Decision
Tree Algorithm. In 2021 IEEE International Conference on Internet of Things and
Intelligence Systems (IoTaIS), 2021, pp. 196-202. IEEE.
[4] A. Jaribion, S. H. Khajavi, J. Holmstrom, “loT-enabled workplaces: a case study of
energy management and data analytics,” IECON 2019 - 45th Annual Conference
of the IEEE Industrial Electronics Society, 2019, pp. 5325-5330, DOI: 10.1109/
IECON.2019.8927003.
[5] Ahvenniemi, H., Huovila, A., Pinto-Seppa, I., Airaksinen, M. “What are the differences
between sustainable and smart cities?” Cities, 2017, 60, 234-245.
[6] Kramers, A., Hojer, M., Lovehagen, N., Wangel, J. “Smart sustainable cities -
Exploring ICT solutions for reduced energy use in cities,” Environmental Modelling &
Software, 2014, 56, 52-62.
[7] Hoang, A.T., Pham, V.V., Nguyen, X.P. “Integrating renewable sources into energy
system for smart city as a sagacious strategy towards clean and sustainable process,”
Journal of Cleaner Production, 2021, 305, 127161.
[8] Adhikari, M., Ghimire, L.P., Kim, Y., Aryal, P., Khadka, S.B. “Identification and ana­
lysis of barriers against electric vehicle use,” Sustainability, 2020, 12, 4850.
[9] Sanguesa, J.A., Torres-Sanz, V., Garrido, P., Martinez, F.J., Marquez-Barja, J.M.
“A review on electric vehicles: technologies and challenges,” Smart Cities, 2021, 4,
372-404.
[10] Omitaomu, O.A., Niu, H. “Artificial intelligence techniques in smart grid: a survey,”
Smart Cities, 2021, 4, 548-568.
[11] Yigitcanlar, T., Mehmood, R., Corchado, J.M. “Green artificial intelligence: towards
an efficient, sustainable and equitable technology for smart cities and futures,”
Sustainability, 2021, 13, 8952.
[12] Crossley, P., Beviz, A. “Smart energy systems: transitioning renewables onto the grid,”
Renewable Energy Focus, 2010, 11, 54-59.
176 IoT and Machine Learning for Smart Applications

[13] Lund, H., Andersen, A.N., 0stergaard, P.A., Mathiesen, B.V., Connolly, D. “From
electricity smart grids to smart energy systems—A market operation based approach
and understanding,” Energy, 2012, 42, 96-102.
[14] Blackmon, D. Renewables Won’t Save Us If the Electric Grid Is Not Ready. Forbes. 2020.
[15] O'Connell, N., Pinson, P., Madsen, H., O'Malley, M. “Benefits and challenges of
electrical demand response: A critical review,” Renewable and Sustainable Energy
Reviews, 2014, 39, 686-699.
[16] Lu, R., Hong, S.H., Zhang, X. “A dynamic pricing demand response algorithm
for smart grid: reinforcement learning approach,” Applied Energy, 2018, 220,
220-230.
[17] Boza, P., Evgeniou, T. “Artificial intelligence to support the integration of vari­
able renewable energy sources to the power system,” Applied Energy, 2021, 290,
116754.
[18] Hashem, I.A.T., Chang, V., Anuar, N.B., Adewole, K., Yaqoob, I., Gani, A., Ahmed, E.,
Chiroma, H. “The role of big data in smart city,” International Journal of Information
Management, 2016, 36, 748-758.
[19] Koot, M., Wijnhoven, F. “Usage impact on data center electricity needs: a system
dynamic forecasting model,” Applied Energy, 2021, 291, 116798.
[20] Crawford, K. The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial
Intelligence. New Haven, CT: Yale University Press, 2021.
[21] Mytton, D. “Data centre water consumption,” npj Clean Water, 2021, 4, 11.
[22] Tiefenbeck, V., Worner, A., Schob, S., Fleisch, E., Staake, T. “Real-time feedback
promotes energy conservation in the absence of volunteer selection bias and monetary
incentives,” Nature Energy, 2019, 4, 35-41.
[23] Li, Y., Wen, Y., Tao, D., Guan, K. “Transforming cooling optimization for green data
center via deep reinforcement learning,” IEEE Transactions Cybernetics, 2020, 50,
2002-2013.
[24] Google. Our third decade of climate action: realizing a carbon-free future. 14
September 2020. Available online: https://blog.google/outreach-initiatives/sustain
ability/our-third-decade-climate-action-realizing-carbon-free-future/ (accessed on 20
December 2022).
[25] Wong, L.A., Ramachandaramurthy, V.K., Walker, S.L., Taylor, P., Sanjari, M.J.
“Optimal placement and sizing of battery energy storage system for losses reduc­
tion using whale optimization algorithm,” Journal of Energy Storage, 2019, 26,
100892.
[26] Almulaiki, W.A. “The impact of performance management on employee perform­
ance,” Saudi Journal of Business and Management Studies, 8(2), 2023, pp. 22-27.
[27] Ikonne Chinyere, Prof. Ikonne Chiemela, “Workplace E-monitoring and surveillance
of employees: Indirect tool of information gathering,” International Journal of Science
and Research (IJSR), 3, no. 10, October 2014.
[28] Vilde Hoff Bernstr0m, Helge Svare, “Significance of monitoring and control for
employees’ felt trust, motivation, and mastery,” Nordic Journal of Working Life
Studies, 7, no. 4, December 2017.
[29] Jeske, D., Kapasi, I. “Electronic performance monitoring: Lessons from the Past and
Future Challenges,” In Organizing for digital economy: Societies, communities and
individuals. Proceedings of the 14th annual conference of the Italian chapter of the
AIS, 2018, pp. 119-132.
Smart Workspace Automation 177

[30] Ahmad, Z., Ong, T.S., Liew, T.H., Norhashim, M. “Security monitoring and information
security assurance behaviour among employees: An empirical analysis,” Information
& Computer Security, vol. 27, no. 2, 2019, pp. 165-188.
[31] Vujovic V, Maksimovic M. “Raspberry Pi as a sensor web node for home automation,”
Computers & Electrical Engineering, 44, 2015, 153-171.
[32] George, A.S., George, A.H., Baskar, T. “The evolution of smart factories: How Industry
5.0 is revolutionizing manufacturing,” Partners Universal Innovative Research
Publication, vol. 1, no. 1, 2023, pp. 33-53.
 Application of Digital
11 Image Watermarking
in the Internet of Things
and Machine Learning
K. Prabha and I. Shatheesh Sam

11.1 INTRODUCTION
In an increasingly connected world, the convergence of advanced technologies
such as Digital Image Watermarking, the Internet of Things (IoT), and machine
learning has opened up new avenues for securing and interacting with digital
content. Digital Image watermarking is the technique that embeds invisible data
(watermarks) in digital media, such as text, images, audio, and videos, in order to
safeguard its ownership, validity, and integrity. This has significant applications
in the fields of machine learning and the IoT. This powerful combination helps us
to address critical issues such as data security, authentication, and robust image
processing.
The IoT, a network of interconnected devices that communicate seamlessly,
generates an enormous amount of data every day in this fast growing environment.
Frequently, this data consists of sensitive visual information, such as surveillance
camera footage, medical images, or industrial sensor data. The incorporation of Digital
Image Watermarking provides a robust solution for data protection and authentica­
tion. Watermarked images can be reliably traced back to their source, ensuring data
integrity and authenticity throughout the IoT ecosystem.
Furthermore, the integration of machine learning techniques enhances the cap­
abilities of Digital Image Watermarking within IoT environments. Machine learning
algorithms help in the detection and extraction of watermarks from images, enab­
ling real-time monitoring and analysis. This synergy can prove invaluable in various
applications (Hussain et al., 2020), from fraud detection in smart cities to quality
control in industrial IoT (IIoT) setups (Shah et al., 2022).
The landscape exploration of Digital Image Watermarking in the IoT and ML
will delve into the pivotal role in enhancing data security, ensuring authenticity, and
facilitating advanced image processing. This journey will reveal the potential of this
combination, shedding light on how it is reshaping the way to safeguard and leverage
digital imagery in our increasingly interconnected world.

178 DOI: 10.1201/9781032623276-11

Application of Digital Image Watermarking in IoT and ML 179

The contribution of this chapter includes the advantages, disadvantages, and

applications of Digital Image Watermarking in IoT and ML. In this chapter, section
2 elaborates the advantages and disadvantages of Digital Image Watermarking in
IoT and ML, section 3 illustrates the applications of Digital Image Watermarking
in the IoT and machine learning, and section 4 concludes the chapter by stating the
importance of digital image watermarking in pointing out the emerging challenges
and opportunities in IoT and machine learning.

11.2 ADVANTAGES AND DISADVANTAGES OF DIGITAL

IMAGE WATERMARKING IN THE PERSPECTIVE OF BOTH
IOT AND ML
There are several advantages of using digital image watermarking in the context of
IoT and ML even though it may pose some challenges.

Advantages:

Copyright Protection: Watermarking helps to protect the intellectual property of

digital images, ensuring proper attribution, and preventing unauthorized use (Begum
& Uddin, 2020).

Authentication: Watermarking can serve as a means of image authentication, image

integrity verification as well as tamper protection.

Traceability: Watermarks can embed data about the source or ownership of images,
facilitating traceability, and accountability in data usage.

Security in IoT: Watermarking helps to secure communication in IoT applications by

confirming the authenticity of images exchanged between devices.

Improved ML Model Robustness: Watermarking can enhance the robustness of ML

models by providing additional features or metadata, resulting in better recognition
and classification.

Disadvantages:

Quality Degradation: Some watermarking techniques may result in a loss of image

quality, impacting the visual appeal and potentially affecting the performance of ML
models.

Vulnerability to Attacks: Watermarking may be vulnerable to certain attacks, such

as removal or modification, compromising the security and integrity of embedded
information.
180 IoT and Machine Learning for Smart Applications

Computational Overhead: Implementing robust watermarking techniques

can introduce computational overhead, especially in resource-constrained IoT
devices.

Limited Standardization: The lack of standardized watermarking techniques can

lead to interoperability issues, making it challenging for widespread adoption in IoT
and ML ecosystems.

Complex Integration in IoT: Integrating watermarking in IoT systems may be

complex due to different devices, protocols, and communication channels, requiring
careful consideration.

Potential Legal Concerns: Depending upon jurisdiction, the use of digital

watermarks may have legal implications, understanding, and complying with rele­
vant laws is essential.

Ethical Considerations: Ethical concerns associated with privacy and consent may
arise, especially in IoT applications, where images may be captured and shared
without the explicit consent of individuals.
Digital image watermarking brings several advantages, including copyright pro­
tection and enhanced security; there are some drawbacks such as potential quality
degradation, susceptibility to attacks, and the necessity for careful integration in
diverse IoT environments. It is vital to balance these factors and choose appropriate
watermarking techniques based on specific use cases and requirements.

FIGURE 11.1 Applications of ML and IoT-based watermarking.

Application of Digital Image Watermarking in IoT and ML 181

11.3 APPLICATIONS
When it comes to the IoT and machine learning (ML), digital image watermarking
has numerous valuable applications. Figure 1 shows the Applications of ML and IoT
based digital image watermarking.

Secure Data Transmission in IoT

Watermarking can embed information like timestamps, device IDs, or encryption
keys into images generated by IoT devices. This guarantees the authenticity of the
data when transmitted to a central server.

Image Authentication in IoT

In IoT applications such as surveillance cameras, watermarks help to verify the authen­
ticity of captured images or videos (Kamili et al., 2021). This supports tampering pre­
vention with critical visual data.

Copyright Protection
In IoT scenarios where images are shared and exchanged, digital watermarking aids
to protect the intellectual property rights of content creators and owners.

Inventory and Product Authentication

Watermarking can be applied to product images in an IoT-based inventory manage­
ment system. This helps in authenticating products, detecting counterfeit goods, and
ensuring data integrity about items in stock.

Medical Imaging and Patient Data

In the context of IoT healthcare applications (Al-kahtani et al., 2022), medical images
can be watermarked to maintain data integrity, protect patient privacy, and validate
the source of the images.

Data Augmentation in ML
In machine learning models, large datasets are often required for training.
Watermarking can be used to create synthetic datasets with known labels, facilitating
model training and improving its ability to generalize across different scenarios.
For example, watermarking can be applied to images to create synthetic data with
established ground truth labels, which can enhance model robustness.

Model Robustness Testing

In ML, watermarking can be used to verify the robustness of image recognition
models against common attacks. Watermarked images can be used as part of adver­
sarial testing to evaluate model performance.

Anti-Plagiarism in ML-based Content Generation

In ML, applications such as text-to-image synthesis or image generation, watermarks
can be applied to ensure that generated images or content are not plagiarized from
external sources.
182 IoT and Machine Learning for Smart Applications

Content Attribution in Social IoT

In IoT systems that share images and multimedia content on social platforms, digital
watermarking can help attribute content to its original creators or sources.

Forensic Analysis
Watermarking can help in the forensic analysis of images and videos collected by
IoT devices, helping law enforcement agencies and investigators in solving crimes.
It’s important to implement robust and secure watermarking techniques to withstand
various attacks, including image manipulation and removal attempts. Additionally,
compliance with privacy regulations and intellectual property rights is crucial when
implementing watermarking in IoT and ML applications.

Improved Data Integrity in IoT

IoT devices often capture and transmit image data in various applications, such as
environmental monitoring, agriculture, or industrial automation. By embedding
watermarks into these images, data integrity (Zhang et al., 2017) can be maintained,
and any unauthorized alterations can be detected. This confirms that decisions made
based on IoT data are reliable.

Real-time Authentication in IoT

In real-time IoT scenarios (Behnia et al., 2019), like autonomous vehicles, drones, or
remote monitoring systems, image watermarking can serve as a quick and efficient
method for verifying the authenticity of images and videos, helping prevent security
breaches, and ensuring trust in the data.

Secure Supply Chain Management

In IoT-based supply chain management, watermarking product images or QR codes
can be used to authenticate the authenticity of products and their origins. This is par­
ticularly useful for combating counterfeiting and ensuring product traceability.

Enhancing Robustness in ML Models

Watermarked images can be used for testing the robustness of ML models against
adversarial attacks. This method assures that ML models perform reliably in the
environment of potential threats or manipulation attempts.

Ethical AI and Content Generation

In AI applications, watermarking can serve as a tool for ethical AI practices because
generated content, such as text or images, is shared publicly. It can attribute AI-
generated content to its source, promoting transparency and accountability.

Privacy-Preserving Medical Imaging

In medical IoT, where patient data is shared for diagnosis and treatment, watermarking
can be applied to protect patient privacy while allowing data to be exchanged securely
among healthcare providers and institutions.
Application of Digital Image Watermarking in IoT and ML 183

Intellectual Property Protection

Watermarking can safeguard the intellectual property of content creators and artists
in IoT applications where images or multimedia content are distributed, shared,
or sold.

Anomaly Detection in Security IoT

Watermarking can be applied to surveillance camera feeds in security IoT to detect
anomalies or unauthorized access. Any tampering with the watermarked images can
trigger alerts for further investigation.

Historical Data Integrity in ML

For historical datasets used in ML research or analysis, watermarking is applied to
maintain the integrity of the information and ensure that it remains unaltered over
time, preserving the dataset’s reliability for future studies.
When implementing digital image watermarking in IoT and machine learning,
it’s crucial to consider factors such as the robustness of the watermarking tech­
nique, computational efficiency, and the precise requirements of the application to
ensure its effectiveness. Additionally, staying updated with the latest advancements
in watermarking technology and security practices is essential to address evolving
threats.

Content Provenance in Social Media and IoT

In the age of social media and user-generated content, digital image watermarking
can help establish the origin and ownership of shared images and videos. This is
particularly valuable in IoT applications where users contribute multimedia content,
ensuring proper attribution and intellectual property protection.

Geo-tagged Image Verification in IoT

For IoT devices with built-in geolocation capabilities, watermarking can include
geographic coordinates and timestamps in images. This assists in verifying the
authenticity of data gathered from multiple locations, making it useful in environ­
mental monitoring and geospatial applications.

Preventing Data Manipulation in Autonomous Systems

In autonomous systems like self-driving cars and drones, digital image watermarking
helps to protect the integrity of sensor data. Any attempts to manipulate sensor­
generated images can be detected through watermark verification, enhancing safety
and reliability.

Securing Data Marketplaces

In IoT ecosystems where data marketplaces facilitate data exchange, watermarking
can be used to ensure the trustworthiness and authenticity of data sources, promoting
fair and secure data transactions.
184 IoT and Machine Learning for Smart Applications

Enhanced Anomaly Detection in ML

Machine learning models, when trained on watermarked datasets containing known
anomalies, can be more effective at identifying unusual patterns or outliers in various
applications such as fraud detection, cybersecurity, and fault prediction.

Watermarking for Compliance and Auditing

Watermarking IoT-generated images can be essential for compliance with industry
regulations and auditing requirements. This affords a mechanism to prove the authen­
ticity of data for regulatory purposes.

Protecting Data Privacy in Surveillance

In IoT-based surveillance systems, watermarks can be applied to protect the privacy
of individuals captured in images and videos. Sensitive information, such as faces or
license plates, can be obscured with watermarks to comply with privacy regulations.

Digital Evidence in Legal Proceedings

Watermarked images and videos captured by IoT devices can be digital evidence in
legal cases. These watermarks act as a tamper-evident seal, ensuring the admissibility
and integrity of the evidence in court.

Enhancing User Trust in AI-Generated Content

As AI-generated content becomes more prevalent, watermarking can be employed
to indicate that the content is machine-generated, fostering transparency, and trust
between users and AI systems.

Securing Remote Sensing Data

In remote sensing applications like satellite imagery and weather monitoring,
watermarking can help safeguard data integrity and authenticity, particularly when
used for critical decision-making processes.
These applications demonstrate the versatility and importance of digital image
watermarking in ensuring data integrity, authenticity, and security in IoT and machine
learning contexts. Choosing the right watermarking technique and parameters
depends on the specific requirements of each application, including the level of
security needed and potential threats that may be encountered.

Data Fusion and Sensor Networks in IoT

In IoT environments with multiple sensors and data sources, watermarking can be
used to incorporate data from diverse sensors while ensuring the authenticity and
alignment of the information. This is essential for applications like smart cities, where
data from various sources, such as traffic cameras and weather stations, are combined.

Adaptive Watermarking in IoT

Adaptive watermarking techniques can be employed in IoT to dynamically adjust
watermark strength or presence based on the data’s sensitivity. For instance, critical
data may have stronger watermarks to protect against tampering, while less sensitive
data may have lighter watermarks to save bandwidth.
Application of Digital Image Watermarking in IoT and ML 185

Visual Search and Content Retrieval in ML

In image-based search and content retrieval systems powered by ML, watermarks can
serve as metadata tags, aiding in precise content retrieval and categorization.

Insider Threat Detection in IoT

Watermarking can assist in detecting insider threats within IoT networks. Any
unauthorized access or data tampering attempts by individuals with access privileges
can be identified through watermark verification.

Protecting Training Data in Federated Learning

In federated learning settings, where ML models are trained across decentralized
IoT devices, watermarks can protect the integrity and privacy of local data on these
devices. This confirms that sensitive information remains confidential while contrib­
uting to the training process.

Anomaly Detection in Medical Imaging

In medical IoT, watermarking can be applied to medical images to detect anomalies,
such as tumors or fractures. This assists healthcare professionals in diagnosis and
monitoring.

Metadata Preservation in IoT Data Streams

Alongside watermarking, IoT devices can embed essential metadata into images,
including device ID, location, and timestamp. Watermarking helps ensure this meta­
data remains intact throughout data transmission and processing.

Ensuring Fairness in ML Algorithms

Watermarking can add fairness indicators to datasets used in ML algorithms. This
guarantees that the training data is well-balanced and represents diverse demo­
graphics, mitigating biases in AI systems.

Data Provenance in Environmental Monitoring

In IoT applications related to environmental monitoring, such as climate studies
or pollution tracking, watermarking can be employed to establish data proven­
ance, assuring that data gathered from various sensors remains reliable for scientific
analysis.

Defense Against Deepfake Attacks

As deepfake technology advances, watermarking plays a vital role in detecting
manipulated or synthetic images and videos. Watermarked content can be a reference
for authentic media.

Supply Chain Transparency in Agriculture IoT

In precision agriculture IoT, watermarked images can track the entire lifecycle of
agricultural products, from planting to distribution. This enhances transparency and
accountability in the food supply chain.
186 IoT and Machine Learning for Smart Applications

Collaborative AI Research
Watermarking is used to protect the intellectual property of research institutions and
organizations involved in collaborative AI research and data sharing. It ensures proper
attribution and credit for contributions.

Quality Control in Manufacturing IoT

Watermarking can be applied to product images in manufacturing processes monitored
by IoT sensors and cameras. This enables real-time quality control by comparing
watermarked reference images with the actual product images, ensuring compliance
with quality standards.

Verification of Remote Operations in Robotics

In remote robotic operations, such as surgery or exploration, watermarking can be
applied to validate the authenticity of video feeds and control signals, safeguarding
against unauthorized access or tampering.

Secure Data Sharing in IoT Ecosystems

Watermarking allows for secure data sharing among IoT devices and stakeholders in
a networked ecosystem. By embedding watermarks in shared images or data streams,
the source and integrity of the data are preserved.

Art Authentication and Provenance

In the art world, digital watermarking can be applied to high-resolution images
of artworks to establish their authenticity and provenance. This is crucial for art
collectors, galleries, and museums.

Financial Document Verification

In the financial sector, digital watermarking can be applied to secure financial
documents, such as checks and certificates. Watermarks can include details like
account numbers, dates, and issuer information, enhancing document security.

Historical Data Preservation in Cultural Heritage

Watermarking can be applied to digitized historical documents, photographs, and
artifacts in cultural heritage preservation efforts. The integrity of these valuable
resources for future generations is maintained by watermark.

IoT-Based Environmental Conservation

In conservation efforts utilizing IoT sensors, watermarked images of wildlife and
ecosystems can help track and protect endangered species and habitats, ensuring the
data’s credibility for research and policy-making.

Personalized Content Generation in ML

ML models can use watermarked images to create personalized content, such as
customized advertisements or product recommendations while ensuring the origin­
ality and authenticity of the generated content.
Application of Digital Image Watermarking in IoT and ML 187

Intellectual Property Verification in 3D Printing

In 3D printing applications, digital watermarking (Begum & Uddin, 2020) is used to
validate the authenticity of 3D models and designs, protecting the intellectual prop­
erty of designers and manufacturers.

Agricultural Pest and Disease Monitoring

In the field of precision agriculture, watermarked images are used to detect pests
and diseases in addition to monitoring crop health. This data is crucial for producing
informed decisions about pesticide use and crop management.

Trustworthy Remote Sensing for Disaster Response

During disaster response efforts that utilize IoT and remote sensing technologies,
watermarked images provide reliable and tamper-evident data for disaster assessment,
response coordination, and decision-making.

Maintaining Ethical AI in Journalism

Watermarking can be applied to AI-generated content used in journalism to ensure
that articles or news reports generated by AI systems are clearly identified as such,
maintaining ethical reporting standards.
These diverse applications underscore the significance of digital image
watermarking in securing, authenticating, and enhancing data across various IoT and
machine learning domains. Implementing the appropriate watermarking techniques
depends on specific use cases and the need to balance data security, privacy, and
usability.

Augmented Reality (AR) and IoT Integration

In AR applications, IoT devices can capture real-world images enriched with digital
overlays. Watermarking can secure these augmented images, guaranteeing that the
virtual elements remain reliable in the physical environment.

Watermarking for Privacy-Preserving AI

Watermarking can help preserve privacy in AI research and applications that
handle sensitive data. For example, facial recognition models can be generated on
watermarked images to avoid unauthorized facial recognition.

Digital Watermarking for Sensor Fusion in IoT

In IoT systems with diverse sensors, watermarks can be used to combine data
from different sensors like cameras, accelerometers, and temperature sensors.
This creates a comprehensive perspective of the environment while preserving
data integrity.

Energy-Efficient Watermarking in IoT

Energy efficiency is crucial in IoT devices. Lightweight watermarking techniques are
designed to minimize computational demands, making them suitable for resource-
constrained IoT devices.
188 IoT and Machine Learning for Smart Applications

Weather Forecasting and Climate Research

Watermarked satellite and sensor data are vital for accurate weather forecasting and
climate research. These watermarks confirm the authenticity of data used for making
critical meteorological predictions.

Enhancing Authentication in ML Model Deployment

When deploying machine learning models in production, watermarked data can
enhance user authentication and authorization processes, safeguarding access to sen­
sitive AI models.

Blockchain and Watermarked Data

Watermarked images and data can be incorporated with blockchain technology in IoT
and ML applications. This combination ensures an immutable and transparent record
of data transactions and model training.

Watermarking for Autonomous Delivery Drones

Drones used in autonomous delivery can capture watermarked images of packages,
facilitating tracking, verification, and proof of delivery in logistics operations.

Remote Learning and Watermarked Educational Content

In online education, watermarked content can help verify the authenticity of educa­
tional materials. It assures students and institutions that the content is legitimate and
from reliable sources.

Watermarking for Product Lifecycle Management (PLM)

In Product Lifecycle Management (PLM) systems used in manufacturing, digital
watermarking can be employed to mark computer-aided design (CAD) designs and
product images at various stages of development, ensuring design integrity and com­
pliance with specifications.

Watermarking in Recommender Systems

Recommender systems in e-commerce (Sherekar et al., 2008) can use watermarked
images and product data to enhance recommendations. This ensures that
recommendations are based on authentic product information.

Compliance Monitoring in Healthcare IoT

In healthcare IoT, watermarked patient data and medical images (Hashim et al., 2020)
can be monitored for compliance with regulations like Health Insurance Portability
and Accountability Act (HIPAA), ensuring patient privacy and data security.

Wildlife Conservation and Watermarked Camera Traps

Camera traps used in wildlife conservation efforts can capture watermarked images
of animals, assisting researchers in population studies and habitat conservation.
Application of Digital Image Watermarking in IoT and ML 189

Watermarking IoT-Generated Art and Creativity

IoT devices can be programmed to generate art or music. Watermarking ensures that
such creative works are attributed to their IoT creators and not plagiarized.
These applications illustrate the versatility of digital image watermarking in
safeguarding data, enhancing security, and preserving authenticity in both IoT and
ML domains. The selection of watermarking technique depends on factors such as
the application’s specific requirements, data sensitivity, and the prospective for mali­
cious tampering.

Watermarking in Autonomous Vehicles

Autonomous vehicles rely heavily on sensor data, including cameras, LiDAR, and
radar. Watermarking can ensure the integrity and authenticity of the data utilized for
navigation, obstacle detection, and decision-making, making autonomous transporta­
tion safer and more reliable.

Agricultural Yield Prediction

In precision agriculture, watermarked images of crops can be used to forecast yields
based on growth patterns, helping farmers to form informed decisions about planting,
harvesting, and resource allocation.

Watermarking in Environmental Monitoring Networks

Environmental monitoring networks often employ a multitude of sensors and cameras
to observe natural phenomena. Watermarking can assist in integrating and securing
data from these sources, ensuring the credibility of environmental research and
decision-making.

E-commerce Product Authenticity

In e-commerce (Rahmati et al., 2013), watermarked product images are used to verify
the authenticity of products listed for sale. This aids buyers in making informed pur­
chasing decisions while protecting them from counterfeit goods.

Watermarking in Human Activity Recognition (HAR)

Human Activity Recognition (HAR) (Muhammad et al., 2021) applications in IoT
and wearables can benefit from watermarking by ensuring the integrity of sensor data
used to recognize human activities such as walking, running, or sleeping.

Watermarking IoT Data for Financial Transactions

In IoT-based payment systems and financial transactions, watermarking can be
applied to transaction records and receipts to prevent fraud and enhance transaction
security.

Watermarking for Authenticating Voice and Sound Data

In addition to images, watermarking can be extended to authenticate voice recordings
and sound data in IoT voice assistants, voice-controlled devices, and audio analytics
applications.
190 IoT and Machine Learning for Smart Applications

Watermarking for Wildlife Tracking

In wildlife conservation efforts, watermarked images and videos from camera traps
can assist in tracking and studying animal behavior, migration patterns, and popula­
tion dynamics.

Protecting IoT-Based Digital Twins

In Industry 4.0 and IoT-driven digital twin simulations of physical assets, watermarking
is used to secure the precision of the digital representation, ensuring it is arranged
with the real-world counterpart.

Watermarking for Preventing Data Poisoning Attacks

In machine learning, watermarking is used to identify data poisoning attacks where
adversaries inject malicious data into training sets. Watermarked data can help iden­
tify and mitigate such attacks.

Maintaining Data Integrity in Remote Sensing for Agriculture

In precision agriculture, remote sensing data is crucial for crop management.
Watermarked data ensures that the imagery used for crop analysis, irrigation planning,
and yield prediction remains unaltered and trustworthy.

Geospatial Watermarking for Navigation

In location-based services and GPS navigation applications, watermarked geospatial
data can enhance location accuracy and trustworthiness, especially in urban canyons
or remote areas.

Watermarking IoT-Generated Artifacts in Industrial IoT (IIoT)

In IIoT settings where machines generate manufacturing instructions,
watermarked digital artifacts can ensure the authenticity and traceability of pro­
duction processes.
These applications emphasize the diverse ways in which digital image
watermarking is instrumental in enhancing data security, reliability, and trustworthi­
ness in the evolving fields of IoT and machine learning. Implementing watermarking
techniques that align with each application’s specific needs and constraints is crucial
for their effectiveness.

Watermarking for IoT-Based Environmental Compliance

Industries and municipalities use IoT sensors for environmental compliance
monitoring. Watermarked data ensures that air and water quality measurements
remain unaltered for regulatory reporting.

Watermarking in Supply Chain Blockchain

In supply chain applications using blockchain, watermarks can be added to product
images and sensor data. This helps in verifying the authenticity and provenance of
products at each stage of the supply chain.
Application of Digital Image Watermarking in IoT and ML 191

Watermarking for Sentiment Analysis

In sentiment analysis, watermarked images with associated emotions are applied to
train ML models. This enhances the accuracy of sentiment predictions in applications
like customer feedback analysis and social media monitoring.

Watermarking for Anti-Counterfeiting in Luxury Goods

In the luxury goods industry, watermarked images of products can be linked to authen­
tication databases, enabling customers to verify the authenticity of their purchases
using a smartphone app.

Watermarking for Remote Inspection in Manufacturing

In remote inspection applications, where IoT devices capture images of machinery
and infrastructure, watermarked images are used to ensure that scrutiny reports are
based on original, unaltered data.

Watermarking for Document Archiving in Legal

Law firms and legal departments can employ watermarking to ensure the integ­
rity of digital documents stored for legal purposes, such as contracts, wills, and
evidence.

Watermarking for Smart Grid Security

In the smart grid, watermarked sensor data ensures the reliability and security
of the energy distribution system, protecting against unauthorized access or
manipulation.

Watermarking for Traffic Management

IoT-based traffic management systems use watermarked images from traffic cameras
to analyze congestion, accidents, and road conditions, aiding in real-time traffic
control.

Watermarking for Authentication in Augmented Reality Museums

In AR-enhanced museum experiences, watermarked historical images and artifacts
can be used to validate the digital overlays, enhancing the educational and immersive
aspects of the exhibits.

Watermarking for Cybersecurity Training Data

In cybersecurity training, watermarked datasets containing simulated attacks and
vulnerabilities can be used to train ML models for intrusion detection and threat
prevention.

Watermarking for Industrial Safety Compliance

In industrial settings, watermarked images can be used for safety compliance
inspections, ensuring that equipment inspections and safety protocols are based on
unaltered data.
192 IoT and Machine Learning for Smart Applications

Watermarking for IoT-Based Wildlife Conservation Databases

Conservation organizations can use watermarked images and data to maintain reliable
databases for tracking wildlife populations and protecting endangered species.

Watermarking in Smart Grid Meter Reading

In smart grid meter readings, watermarked meter images are used to authenticate the
accuracy of energy consumption data, reducing billing disputes and fraud.

Watermarking for Personal Health Monitoring

In wearable health devices, watermarked health data are used for personal health
monitoring and medical diagnoses, ensuring data integrity and privacy.

Watermarking for Drone-Based Infrastructure Inspection

Cameras Drones can capture watermarked images of critical infrastructure such as
bridges and power lines, ensuring the integrity of inspection reports.
These diverse applications demonstrate the critical role that digital image
watermarking plays in safeguarding data, enhancing authenticity, and ensuring the
trustworthiness of information across various sectors, from environmental monitoring
to cybersecurity and cultural preservation. Tailoring watermarking techniques to the
specific requirements of each use case is essential for achieving optimal results.

Data Deduplication and Watermarking

In IoT, where immense amounts of data are generated, watermarking is used to iden­
tify duplicate data entries, aiding in data storage efficiency and ensuring that redun­
dant information is properly labeled.

Watermarking in Edge Computing

In edge computing environments, watermarked data can be processed and verified
locally, reducing the necessity for extensive data transmission and enhancing data
security in IoT applications.

Privacy-Preserving Collaborative ML
Watermarking can be incorporated into federated learning and collaborative ML
approaches, allowing multiple parties to train models without sharing sensitive raw
data. Watermarks can attest to the credibility of contributed data.

Watermarking for Cross-Domain Image Retrieval

In ML-based image retrieval systems, watermarked images are used to improve cross­
domain retrieval accuracy by preserving important metadata and context information.

Watermarking for Image Forensics

Watermarking techniques can be instrumental in image forensics, enabling
investigators to identify tampered or manipulated images in legal and criminal
investigations.
Application of Digital Image Watermarking in IoT and ML 193

Watermarking for Resource-Constrained IoT Devices

Lightweight watermarking algorithms are developed to be efficient and suitable for
resource-constrained IoT devices with limited memory and processing power.

Watermarking for Secure Remote Voting Systems

In remote voting systems, watermarked ballots can help ensure the integrity of votes
and prevent tampering, thus enhancing the security and transparency of the voting
process.

Explainable AI and Watermarking

Watermarked training data is used to improve the interpretability of machine learning
models, contributing to explainable AI by preserving information about critical
features and input variables.

Watermarking for Digital Preservation

In digital libraries and archives, watermarked images and documents can serve as
digital artifacts, preserving cultural heritage and historical records while confirming
their authenticity.

Blockchain-Enabled Watermarking
Combining watermarking with blockchain technology (Wang & Hsu, 2022) offers
a robust solution for data integrity and provenance tracking in diverse applications,
comprising supply chain management and legal documentation.

Privacy-Preserving AI Services
Watermarking can be applied to data used by AI services to confirm that user data
remains confidential and untouched during model training and inference, supporting
privacy-preserving AI.

Watermarking for Monitoring Wildlife Behavior

In wildlife behavior research, watermarked images are used to monitor and track
animal movements, interactions, and habits without disturbing their natural
environment.

Watermarking for Precision Medicine

In personalized medicine, watermarked medical images (Allaf & Kbir, 2019) and
genetic data can facilitate the development of tailored treatments while maintaining
patient data security and authenticity.

Continuous Monitoring of Industrial Equipment

Watermarked images and sensor data from industrial equipment enable continuous
monitoring, ensuring that machinery operates within safe and efficient parameters
and alerting to any anomalies.
194 IoT and Machine Learning for Smart Applications

11.4 CONCLUSION
Digital Image Watermarking emerges as a vital tool in this context, providing robust
mechanisms for data protection, authentication, and ensuring the integrity of visual
content. With its ability to process and analyze vast datasets, machine learning
complements Digital Image Watermarking by enhancing the detection and extrac­
tion of watermarks from images. This synergy opens up new avenues for real-time
monitoring and analysis, fostering innovation in various domains. These advanced
applications underscore the adaptability and importance of digital image watermarking
in addressing evolving challenges and opportunities in IoT and machine learning.
Watermarking techniques should be chosen carefully, considering the specific needs
and constraints of each use case, to achieve optimal results in terms of data security,
authenticity, and trustworthiness. In summary, digital image watermarking finds
extensive applications in IoT and ML, ranging from data integrity and privacy protec­
tion to content attribution and fairness enhancement.

REFERENCES
[1] Al-kahtani, M. S., Khan, F., & Taekeun, W. (2022). Application of Internet of Things
and Sensors in Healthcare. Sensors, 22(15), 5738. https://doi.org/10.3390/s22155738
[2] Allaf, A. H., & Kbir, M. A. (2019). A Review of Digital Watermarking Applications
for Medical Image Exchange Security. In M. Ben Ahmed, A. A. Boudhir, & A. Younes
(Eds.), Innovations in Smart Cities Applications Edition 2 (pp. 472-480). Springer
International Publishing. https://doi.org/10.1007/978-3-030-11196-0_40
[3] Begum, M., & Uddin, M. S. (2020). Digital Image Watermarking Techniques: A
Review. Information, 11(2), 110. https://doi.org/10.3390/info11020110
[4] Behnia, R., Ozmen, M. O., & Yavuz, A. A. (2019). ARIS: Authentication for Real-Time
loT Systems. ICC 2019 - 2019 IEEE International Conference on Communications
(ICC), 1-6. https://doi.org/10.1109/ICC.2019.8761207
[5] Hashim, M. M., Rhaif, S. H., Abdulrazzaq, A. A., Ali, A. H., & Taha, M. S. (2020).
Based on IoT Healthcare Application for Medical Data Authentication: Towards
A New Secure Framework Using Steganography. IOP Conference Series: Materials
Science and Engineering, 881(1), 012120. https://doi.org/10.1088/1757-899X/881/1/
012120
[6] Hussain, F., Hassan, S. A., Hussain, R., & Hossain, E. (2020). Machine Learning for
Resource Management in Cellular and IoT Networks: Potentials, Current Solutions,
and Open Challenges. IEEE Communications Surveys & Tutorials, 22(2), 1251-1275.
https://doi.org/10.1109/COMST.2020.2964534
[7] Kamili, A., Hurrah, N. N., Parah, S. A., Bhat, G. M., & Muhammad, K. (2021).
DWFCAT: Dual Watermarking Framework for Industrial Image Authentication and
Tamper Localization. IEEE Transactions on Industrial Informatics, 17(7), 5108-5117.
https://doi.org/10.1109/TII.2020.3028612
[8] Muhammad, K., Mustaqeem, Ullah, A., Imran, A. S., Sajjad, M., Kiran, M. S., Sannino,
G., & De Albuquerque, V. H. C. (2021). Human Action Recognition Using Attention
Based LSTM Network with Dilated CNN Features. Future Generation Computer
Systems, 125, 820-830. https://doi.org/10.1016/j.future.2021.06.045
[9] Rahmati, P., Adler, A., & Tran, T. (2013). Watermarking in E-commerce. International
Journal of Advanced Computer Science and Applications, 4(6), 256-306. https://doi.
org/10.14569/IJACSA.2013.040634
Application of Digital Image Watermarking in IoT and ML 195

[10] Shah, S., Iqbal, M., Aziz, Z., Rana, T., Khalid, A., Cheah, Y.-N., & Arif, M. (2022). The
Role of Machine Learning and the Internet of Things in Smart Buildings for Energy
Efficiency. Applied Sciences, 12(15), 7882. https://doi.org/10.3390/app12157882
[11] Sherekar, M. S. S., Thakare, D. V. M., & Jain, D. S. (2008). Role of Digital Watermark
in E-governance and E-commerce. International Journal of Computer Science and
Network Security, 8(1), 257-261.
[12] Wang, C.-H., & Hsu, C.-H. (2022). Blockchain of Resource-Efficient Anonymity
Protection with Watermarking for IoT Big Data Market. Cryptography, 6(4), 49.
https://doi.org/10.3390/cryptography6040049
[13] Zhang, G., Kou, L., Zhang, L., Liu, C., Da, Q., & Sun, J. (2017). A New Digital
Watermarking Method for Data Integrity Protection in the Perception Layer of IoT.
Security and Communication Networks, 2017, 1-12. https://doi.org/10.1155/2017/
3126010
Index
A H
Actuators, 21, 24, 45 Honey badger optimization, 113
Augmented reality, 187
I
B
Industry 4.0, 190
Bigdata, 33, 146 Instance-based learning, 5
Blockchain, 97, 174 Intelligent IoT, 33
Bluetooth, 58, 134 Intelligent transportation system, 6
Internet of Things, 127
C Internet of Things in industry, 28, 33
IoET, 174
Cellular networks, 59
Clone block attack, 97
K
Cloud computing, 22, 33
Cobots, 174 K-Means clustering, 74
Context-dependent sentiment classification,
145 L
Convolutional neural network, 145
Linear regression, 153
Cutting edge technologies, 130
Long Term Evolution Advanced (LTE-A), 135
Cybersecurity, 5
Long-Term Evolution (LTE), 135
Low-power WAN, 58
D
Database as a Service (DBaaS), 47 M
Data heterogeneity, 70
Machine-to-Machine (M2M) communication, 33
Decision trees, 71
Mesh networking, 59
Digital watermarking, 187
Microcontrollers, 24, 45
Mobile edge computing, 97
E
MQTT, 166
Edge computing, 22, 35, 59 MultinomialNB, 146
Embedded systems, 25 Multipliers, 112
Encircling, 104 Multitask learning, 4
Ensemble learning, 4
N
F
NB-IoT, 58
Feedforward neural network, 145 Network functions virtualization, 136
Field programmable gate array, 111 Neural network learning, 4
5G Connectivity, 33 NLP, 113

G P
Global roaming, 60 Perception layer, 40
Grey wolf optimization, 103 Principal component analysis, 145

197
198 Index

Q T
Quantum key distribution, 136 Three-layer architecture, 40
Quantum networking, 136 Title firmness, 42
Transfer learning, 105
R
U
Random forest, 73
Reinforcement learning, 3, 82 Ultra-wideband, 134
Renewable energy, 7 Unsupervised learning, 2, 81
RFID, 33, 133, 164
V
S
VGG16, 107
Semi-supervised learning, 3 VGG16-TL, 102
Sensing and Actuation as a Service (SAaaS),
47 W
Sensing as a Service (SaaS), 47
Wearable devices, 27
Sensor as a Service (SenaaS), 47
Wireless fidelity, 134
Sensors, 21, 24, 45
Wireless sensor networks, 134
Sigfox, 58
6LoWPAN, 43
Z
Smart grid, 130
Supervised learning, 2, 81 ZigBee, 42, 58, 134
Support vector machines, 71, 115 Z-wave, 58