Prediction using Neural Networks and other eCommerce platforms use Al to show the related and

Neural Networks recommended products. The compilation is formed on the basis of the
 A neural network is a type of machine learning based on the users' behavior. The system analyzes the characteristics of certain it.
model of the human nervous system using a series of 9. Logistics
mathematical and physical processes. Neural networks can be This industry needs a lot of management that is to be done manually by
designed to perform tasks such as decision-making. employees of many companies. But nowadays, neural networks are capable
 A series of algorithms that endeavors to recognize underlying of routing and dispatching.
relationships in a set of data through a process that mimics the For example, Wise Systems is an autonomous system which lets a user:
way the human brain operates plan routes and monitor them; customize shipping routes in real-time with
ROLE OF NEURAL NETWORKS IN PREDICTIVE ANALYTICS the help of predictive features.
 Effective predictive analytics requires more than simply.
Identifying and analysing past data. It also requires that you Chapter 2: CHAID (Chi-square Automatic Interaction Detector)
uncover previously unknown relationships between your What is CHAID (Chi-square Automatic Interaction Detector) analysis?
existing data and the future outcomes you want to predict.  is an algorithm used for discovering relationships between a
 This is where artificial neural networks come in. Neural categorical response variable and other categorical predictor
networks are self-learning systems that use algorithms to variables.
analyze large amounts of data. They can then make predictions  It's based on signif cance testing and came into use during the
based on those relationships they remembered from the self- 1980s after Gordon Kass published An Exploratory Technique
learning systems. for Investigating Large Quantities of Categorical Data.
 An artificial neural network (ANN), often simply referred to as a It involves:
neural network (NN), is an attempt to replicate the brain's Regression: A statistical analysis to estimate the relationships between a
network of neurons so that the computer may learn and make dependent response variable and other independent ones.
judgments in a human-like manner. Simply put, they are Machine learning: Artificial intelligence that leverages data to absorb
computational models, or what we know as algorithms. information, then makes logical predictions or decisions.
Structure of A Neural Network Prediction Decision trees: Branching models of decisions or attributes, followed by
The input layer of a neural network is where you will enter your data for their event outcomes.
analysis. Purposes for using CHAID in a B2B research project include:
The output layer of a neural network is where the results will be displayed. • Refining buyer personas and segments
The hidden layers make neural networks so powerful they are, in essence, • Understanding data behind brand perceptions
the "brain" of the neural network, where complex calculations take place to • Informing product or service features
determine the outcome at the output layer. • Exploring customer reactions to marketing
3 Layers of Neural Network Prediction • Predicting and influencing the buying process
1. Input Layer CHAID use cases
2. Hidden Layer Market segmentation: Statistical analysis techniques such as CHAID can help
3. Output Layer you identify segments. in your customer base. Understanding the variety of
THE CAPABILITIES OF NEURAL NETWORKS traits, wants, needs, and behaviors in your target audience should inform a
Adaptive leaming tailored approach to sales and marketing.
Neural networks, like humans, represent non-linear and complicated Perception/brand tracking: Understanding the different segments for
interactions, as well as building on previous knowledge. Software, for specific brand perceptions, plus these groups' notable characteristics. For
example, uses adaptive learning to educate children in math and language example, you could use CHAID to analyze different groups customer
arts. satisfaction or likelihood to recommend.
Self-organization New product development: Testing a detailed concept with fleshed-out
The ability to group and classify massive amounts of data makes neural features using CHAID can support product development. A decision tree can
networks ideal for dealing with the complex visual issues presented by show how different groups would respond to a product concept and reveal
medical imaging these images are often difficult for humans to decipher, the impact of different characteristics. Additionally, you could use CHAID to
whereas a neural network can learn to group the different body part. evaluate how customers would choose between multiple concepts.
Pattern Recognition Marcomms testing. As with product testing, you could test how customer
Neural networks excel at the challenge of identifying faces by learning to appeal for different messages varies, or how different segments would
identify patterns in facial features and expressions. This talent makes them respond to a single marketing message.
perfect for applications such as security systems that must analyze live Advantages of using CHAID
video footage. Reading between the lines; By scientifically evaluating the relationship
Real-time operation between variables, you could get more accurate results compared to asking
Neural networks can, on occasion, provide real- time responses, as can be respondents directly. Sometimes, respondents do not know exactly why
seen with self- driving cars and drone navigation. they make certain decisions and cannot pinpoint the driving factors
Big Data analysis themselves. Detailed outputs: CHAID's decision trees can be broad e.g. with
Neural networks can provide valuable assistance when analyzing large several long branches and lots of leaves. In contrast, some other decision
datasets. The algorithms used by neural networks help to reveal relevant tree techniques are binary i.e. they only produce two long branches
patterns. and relationships between variables, which may not be obvious showing a couple of outcomes.
using other data-analysis tools. Simple to understand and share: Decision trees are more visual than, for
Applications of Neural Networks example, data tables. They have a logical flow and show results in a more
1. Facial Recognition self-explanatory way than many other statistical techniques. Therefore, the
Facial Recognition Systems are serving as robust systems of surveillance. results are easier to share with different teams across your business.
Recognition Systems matches the human face and compares it with the A CHAID diagram typically includes:
digital images. They are used in offices for selective entries. The systems Root: This is the starting point for the decision tree, with lines a.k.a.
thus authenticate a human face and match it up with the list of IDs that are branches stemming from it. In market research, the root shows the overall
present in its database. factor used to find contrasting segments.
2. Stock Market Prediction Branches: Branches connect the nodes, also known as leaves. In research,
Investments are subject to market risks. It is nearly impossible to predict the where the prediction model identifies several customer segments, there will
upcoming changes in the highly volatile stock market. be many branches - each with its own leaves.
3. Customer segmentation Leaves: Leaves show the criteria included in the final prediction model. In
One of the most important marketing tools is customer segmentation. By market research analysis, leaves are the variables or traits found in a
dividing customers into groups based on various characteristics, a company specific customer segment.
can better target its marketing efforts and create products that meet example
specific needs. Neural networks are excellent at grouping customers into CHAID is often used in direct marketing to understand how different groups
segments; they can even do so when there are no apparent similarities of customers might respond to a campaign based on their characteristics.
among the group members. for example, that we run a marketing campaign and are interested in
4. Customer response modelling understanding what customer characteristics (e.g., gender, socio-economic
Another important aspect of marketing is understanding how customers status, geographic location, etc.) etc.) are are associated associat with the
respond to specific strategies and products. You can train neural networks response rate achieved. We build a CHAID "tree" showing the effects of
to predict customer behavior based on previous history and other. relevant different customer characteristics on the likelihood of response.
information. This allows businesses to test their strategies before they are Using CHAID for data analysis: best practices in B2B
implemented, leading to more. 1. Prune the decision trees
5. Fraud detection  Some human input helps shape the analysis to maximize its
By using neural networks to detect fraudulent transactions, companies in usefulness. In CHAID, that often requires some 'pruning' of
every industry can save money and increase efficiency. Neural networks your decision tree outputs.
outperform other. fraud detection methods because they can learn from  Your initial decision tree results may include leaves that are
historical data; this ensures that the system adapts as new trends emerge relatively less important to stakeholders, for example.
over time.  By taking these out of the equation, you can simplify a tree -
6. Sales forecasting making it simpler to understand and more useful for meeting
Predictive neural networks can be used to analyze sales data and make your research objectives.
accurate predictions about future sales trends, enabling businesses to 2. Prioritize the big picture over methodological detail
optimize inventory, production, and pricing strategies. • Setting up a CHAID analysis can be complex. Therefore, it’s easy
7. Marketing optimization and common to debate small tweaks to parameters that could
Predictive neural networks can be used to analyze. customer data and alter the research methodology or approach.
predict future buying behavior, enabling businesses to optimize marketing • Ultimately, minor changes tend to have little impact on the
campaigns, improve customer engagement, and increase sales. overall results. But if discussed in detail, they can have a major
8. Commerce impact on project timelines and slow down the process.
This technology is used in this industry for various purposes. But the most • Prioritize the end goal of a statistical market simulation, rather
frequent example of artificial neural network application in eCommerce is than getting every precise methodological detail right. CHAID is
personalizing the purchaser's experience. For instance, Amazon, AliExpress, a quick technique and you can re-run it if necessary.
3. Use a robust sample size for instance ‘home’ becomes ‘house’. Lemmatization considers
• CHAID analysis requires a large sample size, to split customers the context of the word used, as the same word can have
into multiple groups and test for a significant difference different meanings depending on where and how it’s used.
between them. The following are some areas in which NLP is proving to be extremely
• It means that quantitative market research projects with a low helpful in analyzing unstructured text data:
sample size won’t get good results from CHAID. • Financial analysis: Financial analysts and traders can make
CHAID uses Person’s chi-squared test of independence, which test for an informed decisions by analyzing news about companies. NLP is
association between two categorical variables. A statistically significant used to track news, such as reports of possible mergers,
result indicates that the two variables are not independent, i.e., there is a financial results, etc., and incorporates the analysis into trading
relationship between them. algorithms.
• It works with categorical target variable “Success” or “Failure”. • Recruitment: NLP can help to extract information about
• It can perform two or more splits. different candidates from their cover letters, emails and
• Higher the value of Chi-Square higher the statistical significance resumes that are in widely diverse formats.
of differences between sub-node and Parent node. 2. Image recognition identifies objects, people, and scenes within images. It
• Chi-Square of each node is calculated using formula, is highly beneficial for analyzing visual data such as photographs and
• Chi-square = ((Actual — Expected)² / Expected)¹/2 illustrations. Image recognition techniques such as object detection enable
• It generates tree called CHAID (Chi-square Automatic organizations to recognize user-generated content, analyze product images,
Interaction Detector) and extract texts from scanned documents for further analysis.
3. Video analytics involves extracting significant information from video
data, such as identifying patterns, objects, or activities within the footage.
Chapter 3: Analysis of Unstructured data This technology can serve numerous purposes, including security and
What is unstructured data? surveillance, customer behavior analysis, and quality control in
• Unstructured data are datasets that have not been structured manufacturing. Techniques, such as motion detection, object tracking, and
in a predefined manner. Unstructured data is typically textual, activity recognition, enable organizations to gain insights into their
like open-ended survey responses and social media operations, customers, and potential threats.
conversations, but can also be non-textual, like images, video, 4. Audio analysis tools can process and analyze audio data, including voice
recordings, music, and environmental sounds, to extract useful information
and audio.
or identify patterns. Audio analysis techniques, such as speech recognition,
emotion detection, and speaker identification, are used in multiple
industries like entertainment (content generation, music recommendation),
customer service (call center analytics, voice assistants), and security (voice
biometrics, acoustic event detection).
Best practices on how to make the most of unstructured data
• Develop a clear data strategy. Define your organization's
objectives and requirements for unstructured data analysis.
Text documents. You may encounter unstructured data in the form of text
Identify the data sources, types of analytics to be performed,
documents, which can be plain text files (.txt), Microsoft Word documents
and the expected outcomes to guide your efforts.
(.doc, .docx), PDF files (.pdf), HTML files (.html), and other word processing
• Build data architecture. To effectively leverage unstructured
formats. They primarily contain written content and may include elements
data, allocate resources for creating a comprehensive data
like text, tables, and images.
architecture that supports the storage, management, and
Emails. As a form of electronic communication, emails often contain
analysis of various data types.
unstructured text data and various file attachments, such as images,
• Choose the right tools and platforms. Following the previous
documents, or spreadsheets.
step, you must evaluate and select appropriate tools and
Images. Image files come in various formats, such as JPEG (.jpg, .jpeg), PNG
platforms for unstructured data analytics based on your
(.png), GIF (.gif), TIFF (.tiff), and more. These files store visual information
organization's specific needs, data types, and resources.
and require specialized techniques such as computer vision to analyze and
Consider the scalability, flexibility, and integration capabilities
extract data.
of the solutions.
Audio files. Audio data is usually presented in formats such as MP3 (.mp3),
• Invest in data governance. Establish robust data governance
WAV (.wav), and FLAC (.flac), to name a few. These files contain sonic
policies and processes to ensure data quality, security, and
information that requires audio processing techniques to extract
compliance. Implementing data cataloging, classification, and
meaningful insights.
metadata management can facilitate easier access and retrieval
Video files. Video data comes in popular formats such as MP4 (.mp4), AVI
of unstructured data, enabling more thorough analysis.
(.avi), MOV (.mov), and others. Analyzing videos requires combining
• Build a skilled analytics team. Assembling a multidisciplinary
computer vision and audio processing techniques since they contain visual
team with expertise in data science, machine learning, and
and auditory information.
domain knowledge is a must, as such a team can effectively
Log files. Generated by various systems or applications, log files usually
analyze unstructured data. It is essential to provide training and
contain unstructured text data that can provide insights into system
support to develop their skills and ensure they stay up to date
performance, security, and user behavior.
with industry trends.
Sensor data. Information from sensors embedded in wearable, industrial,
• Foster a data-driven culture. Encouraging a data-driven
and other IoT devices can also be unstructured, including temperature
mindset throughout the organization can be achieved by
readings, GPS coordinates, etc.
promoting data literacy and emphasizing the importance of
Social media posts. Data from social media platforms, such as Twitter,
data-driven decision-making. Sharing insights gained from
Facebook, or messaging apps, contains text, images, and other multimedia
unstructured data analysis with relevant stakeholders and
content with no predefined structure to it.
departments can support collaborative decision-making and
two categories of Unstructured data
foster a data-driven culture.
• human-generated unstructured data which includes various
• Pilot and iterate. To ensure the feasibility and effectiveness of
forms of content people create, such as text documents,
unstructured data analytics initiatives, it’s better to begin with
emails, social media posts, images, and videos; and
small-scale pilot projects. Take the knowledge gained from
• machine-generated unstructured data, on the other hand,
these pilots to refine your approach and scale up successful
which is generated by devices and sensors, including log files,
projects to achieve sustained success.
GPS data, Internet of Things (IoT) output, and other telemetry
• Ensure data security and privacy. Implement robust security
information.
measures and adhere to relevant data protection regulations to
Unstructured data analysis
safeguard the privacy and security of your unstructured data.
1. Natural Language Processing (NLP), a subfield of artificial intelligence, is
Keeping data anonymized or pseudonymized when necessary
a technique that facilitates computer understanding, interpretation, and
can help maintain privacy. It is also important to maintain
generation of human language.
transparency with stakeholders regarding data handling
• It is primarily utilized for analyzing text-based unstructured
practices.
data like emails, social media posts, and customer reviews.
• Measure and optimize. Regularly evaluating the performance
• Text classification, a core NLP technique, simplifies text
and impact of unstructured data analytics efforts by tracking
organization and categorization for easier understanding and
relevant metrics and KPIs is crucial. Doing so enables
utilization. This technique enables tasks like labeling
optimization of processes, tools, and techniques to maximize
importance or identifying negative comments in feedback.
the value derived from unstructured data.
Sentiment analysis, a common text classification application,
categorizes text based on the author's feelings, judgments, or
opinions. This allows brands to understand audience
perception, prioritize customer service tasks, and identify
industry trends.
Let us look at some of the models used by NLP to process unstructured text:
• Stop words removal is the process of removing articles and
prepositions such as ‘the’ and ‘to’ from the text as they add no
value to the NLP process. Most analysis starts with the basic
stop words list which is then augmented based on the specific
objective.
• Stemming is an NLP process that removes affixes, or additions
to a root word by way of a prefix before the word or a suffix
after the word. Stemming can help to group the different forms
of a word together, for example, ‘guitar’ and ‘guitarist’.
• Lemmatization is the process of resolving words to their
dictionary form, known as ‘lemma’. For example, tenses are
removed, so ‘teaching’ and ‘taught’ both become ‘teach’.
Synonyms are also unified during the lemmatization process,