INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

A Review On Big Data With Machine Learning

And Fuzzy Logic For Better Decision Making
J. Betty Jane, Dr.E.N.Ganesh

Abstract:-Big data is playing an important role for many years due to the vast and vide growth of data in various fields and it has become difficult to find
the sense from all those large collective data. Since there is an increase in the large amount of data it is difficult to make better decisions .To overcome
these challenges, machine learning and fuzzy logic are the two techniques that are used for getting better knowledge from various and such large
amount of data sources. This paper discuss about the review on knowledge based decision making that are derived from these two techniques and how
efficiently these two techniques provides the decision making in their own way and the steps involved in these techniques for decision making. Machine
learning is a technique of artificial intelligence to give better intelligent decisions. Fuzzy logic is a process of giving conclusion or decisions from an
uncertain problem.

Keywords: bigdata, fuzzy logic, machine learning

——————————  ——————————

1.INTRODUCTION: 2 .BIG DATA PROCESSING:

Nowadays the data acquired from various sources is very Big data is the term used for huge and large datasets where
large that are generated from the modern equipment such the data processing traditional application software’s are
as sensors, smart phones and internet [1]. The smart not able to deal with these data. The term big data emerged
sensors are used to record the activities that are being fixed in the year 2001 and defined by Doug Laney an analyst [5].
in the cities and in some other organizations for finding the Big data is defined as a huge volume, variety and velocity
activities in those areas. The challenging part of the big of data that has challenges such as enhanced and detailed
data is to find the useful information from these large insight and mainly decision making. Big data is classified in
amounts of data for future developments or improvements. to 8V’S such as volume, velocity, veracity, value, variability,
Hence the knowledge in the extraction process has to be volatile, validity. Value means statistical, correlational
efficient and it has to be an approachable technique for the values. Volume includes extra bytes and zeta bytes. The
real time since the storage of the data is high in many fast generation of data and speed flow of data is known as
situations. The HACE theorem states that the huge-volume, velocity. Validity gives correct data and accurate data. The
autonomous sources with distributed and decentralized structured and unstructured data comes under variety.
control and which help in exploring complex and generating Variability is denoted by the variation in the data speed
relationships among the data. Hence these features made a rates. The validation of data comes under volatility. Veracity
challenge for extracting new useful information from these denotes the trustworthiness in the data.
large Sources. There are two techniques involved for data
extraction, in that machine learning tools are used to
develop the domain knowledge through the learning. This
tool helps in successful extraction of needed information by
the learning process [2]. The machine learning involves
various algorithms for discovering hidden features and
patterns from recent trends and data. The ML algorithms
are used in various areas and in detecting spammed and
unspammed mails by introducing it into the ML
algorithms.[3] A large change is been achieved with the
help of fuzzy logic. After the emergence of fuzzy logic, the
work has become easier, saves money, time and energy.
LottiZadeh is the one who proposed fuzzy logic in 1965.
Before fuzzy logic there were, three ,four valued logic which
has helped for the extension of Boolean logic which helps in
accepting only values like true or false(0 or1).The Boolean
logic gives only two values one is 0 or 1.that is low or
high.no other values can be found in this logic. Hence to Fig a:8v’s of big data
find the most accurate values in the scenario we go for
fuzzy logic which gives the extended view such as low, very A. Big data applications:
low, high, very high also. [4]. In this era, the organizations produce large amount of data
in the form of Exabyte and zettabytes. various fields such
________________________ as
1. Healthcare
 Author name: J..betty jane (research scholar,vels 2. Education
university)E-mail: [email protected]
 Co-Author name: Dr.E.N.Ganesh (Dean,vels university)
3. Media
E-mail: [email protected] 4. Information technology

1121
IJSTR©2019
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

go for machine learning where machine learning is nothing

but teaching the machines how to handle the data.[8]
F. Types of machine learning:
The types of machine learning are:
i. Supervised learning, ii. Unsupervised learning
iii.reinforcement learning
i. supervised learning:

Fig b:big data processing

B. Big data challenges:

There are challenges in big data and these challenges are
1. Dealing with the growth of data.
2. Validating data.
3. Securing large volume of data.
4. Highly cost effective.
5. Massive data sets integration. Fig c: Types of Machine learning

C. Big data analysis tools: The algorithms that need external support are called
The data is not useful until it gives Data that is unstructured supervised learning algorithm. The given input dataset is
for the management to do decision making. To enhance the splitted as training data and test data. The trained dataset
decision making by stages we use some software data with output variable needs to be classified and predicted. All
tools for storing and analyzing the data. Some of the tools these learning algorithms find some pattern from the
are Apache Hadoop, cloudera, Cassandra, mango DB etc. training dataset and then apply it for the classification or
1. Apache Hadoop- This framework is used for clustered file prediction process.
system. Hadoop consists of two main components: The
HDFS for the storage big data, and Map Reduce for big
data analytics [6]
2. Cloudera –collection and processing of unlimited data.
3. Cassandra– managing high volumes of data.
4. Mongo DB- MongoDB is written in C, C++, and
JavaScript and used for document-oriented database.

3.MACHINE LEARNING FOR DECISION

MAKING:
The paradigm of learning or understanding the past Fig d: machine learning algorithm processing
experiences and with that data, trying to improve the future
performance is known as machine learning.[7]. There are two types of supervised learning they are
i. regression
D. Why we go for machine learning method for decision ii.Classification
making: Under regression there are three decision making
The machine learning provides alternative solutions from algorithms:
the large amount of vast data by developing algorithms for They are
processing real time data and gives the accurate results a. Decision tree:
and analysis.The machine learning aims on the computer Decision trees are trees that form a group by sorting based
program development that can access the data and they on their given values. For the purpose of classification
can learn on their own. purpose it is used. The decision tree consists of branches
E.ML Algorithms for decision making: and nodes. The branches represent the node value and the
The main advantage of ML algorithms is, once the nodes represent the attributes to be classified.
algorithm it finds or understands how to analyze the data, it
will do its work automatically. In case of large date volume,
human learning will not be that much effective. Hence we

1122
IJSTR©2019
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

Fig g: Naïve bayes algorithm

Fig e: Decision tree e. support vector machine:
b. Linear regression: Support vector machine is a linear division method. The
A machine learning linear regression algorithm is a division is done as two categorical points that divide their
supervised learning model. The performance of this method points into two-dimensional space, SVM is used to map the
is doing the task of regression.Regression provides a target low-dimensional points in the curve to the high dimensional
prediction values on independent variables. It is used to space and hence they are separable linearly.[10]
build and finding out the relationship between prediction
forecasting and variables.

Fig h: support vector machine

f. k-nearest neighbor algorithm:\

The KNN algorithm is an efficient classification algorithm.
Fig f: Linear regression
[11]They mainly adapt three rules:
They are:
c. Logistic regression: Logistic regression is a supervised
Rule: 1 move the k-nearest code vectors, if the code vector
classification algorithm. It is a predictive analyzer. It
matches the sample training set, then the code vector
transforms the output values using the function called
matches the training vector.
sigmoid and return the value of probability.
Rule: 2Only the k’th and the k + l’th nearest vector codes
Under classification, there are three decision algorithms:
are moved
They are
Rule: 3The k + 1nearest code vectors are moved.
d. Naive bayes algorithm:
Bayesian classification algorithm, is a Bayesian theory and
ii. Unsupervised learning:
it is also called as the statistical algorithm. Bayesian
Unsupervised learning is an artificial intelligence algorithm
classifier is developed from a training dataset with labels of
in which the input sets are classified without any proper
classes. Assuming n is an attributes A1, A2,…, An, an
guidance. Under unsupervised learning there are two
instance E is a vector representation , where ai is the Ai
classifications clustering and reduction. Under clustering,
value, C is used as a class variable representation, c is the
the variables are grouped into clusters and they may be
value of C, and c(E) represents the class label E where it
either user defined or model based one.
belongs to. [9].
a. Cluster analysis:
Cluster analysis is the process of grouping the clusters with
similar data. There are different types of clustering such as
c-means, hierarchical, iterative clustering.[12].

1123
IJSTR©2019
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

c. Algorithm for fuzzy decision making:

The steps involved in fuzzy decision making are as follows:
Fig i:cluster analysis i. In the first step is identification of variables and the
alternatives is done.
b. Dimensionality reduction: ii. The fuzzification process involves converting the
The main aim of dimension reduction algorithms is to linguistic ones from real variables.
reduce the noise or irrelevant information of the data.[13] iii. The variables that need to be taken in the knowledge
There are algorithms for dimensionality reduction they are base are selected by the user.
as follows: iv. The membership function expressed in mathematical
Principal Component Analysis (PCA), function is called membership function
Non-negative Matrix Factorization (NMF), v. The next step is to give the if- then condition rule. Each
Random projection (RP) and auto-encoder (AE) variable represents one rule.
vi. The next step is getting output variable by converting the
fuzzy value.
vii. The last phase of fuzzy process is the implementation of
alternative in practice. if the implementation is Successful it
will improve the system operation with respect to the aim of
the process.

Fig j: Dimensionality reduction

iii. Reinforcement learning:

Reinforcement learning is an artificial intelligence branch in Fig k: fuzzy logic decision making
which allows automatically determining the behavior with in
a small context to the machines.
V.COMPARITIVE REVIEW ON FUZZY LOGIC
IV. FUZZY LOGIC FOR DECISION MAKING: AND MACHINE LEARNING FOR DECISION
Fuzzy logic is a "degrees of truth" approach rather than the MAKING:
"true or false" (1 or 0). The idea of fuzzy logic was first Machine learning algorithms aims at extracting knowledge
introduced by Dr. LotfiZadeh from University of California in from large amount of data and gives traditional methods for
the 1960s.Fuzzy logic gives the decisions as same as that classification and clustering process. It handles multi variety
of human perception and reasoning. Fuzzy logic has been of data and they can be used in large environments.
proved to be applied in well expertise systems. The creation Moreover, learning time is needed more for the algorithms
of fuzzy sets help in determining whether the element to perform the progress in accuracy and relevantly. Fuzzy
belongs to the set or not ranging from 0 to 1.It is used for logic measures the certainty of the problem, and the
making the decisions in the terms of uncertainty.[14,15] algorithms are robust and they adapt easily to the changing
environments.
a. Fuzzy sets:
Fuzzy sets are crisp sets where the characteristic function VI.CONCLUSION:
are transformed to the membership function A: X →[0, 1] The machine learning techniques for decision making
produce good results by handling large data environments
b. properties of fuzzy sets: and give good ideas to the experts in different fields to
improve the future enhancements in the fields they are
involved. Fuzzy logic also equally helps in finding out the
uncertainties in the problem and they adapt themselves to
the changing environments and also help in decision
making. In this paper, the review gives the understanding
that machine learning is an emerging field and its

1124
IJSTR©2019
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616

performance is well compared to fuzzy logic. Hence to Kumar, CSI COMMUN I C A T I O N S | N O V E M

improve the fuzzy logic performance we can enhance the BER 2017
fuzzy extensions on ML algorithms.

REFERENCES:
[1] Vınod Jha, K.Parvathi ―Complete Forward and
Reverse Text Transcription of Odia Braille and
Hindi Braille‖, : Progress in Advanced
Computing and Intelligent Engineering,
Springer, sequence no.12,201
[2] ].Machine Learning for Big Data Processing:A
Literature Review:Bharat Kumar Padhi1 , Dr. S
SNayak2 , and Dr. B N Biswal3,December 2018 |
IJIRT | Volume 5 Issue 7 | ISSN: 2349-6002.
[3] A Very Brief Introduction to Machine LearningWith
Applications to Communication Systems,King’s
College London, United Kingdom ,Osvaldo
Simeone,IEEE.
[4] ].Machine learning and its applications: A
review,shhenaangra,sachinahuja,@2017
[5] Application of fuzzy logic: A literature review,
RenuMakkar,International Journal of Statistics and
Applied Mathematics 2018; 3(1): 357-359
[6] ].A Review on Big Data theoretical and Application
approach:Past and Present: RabelGuharoy,
SumitKumar, Upasak Pal;©2017 IEEE.
[7] Big Data Analytics: A Literature Review
PaperNadaElgendy and Ahmed Elragal,Department of
Business Informatics & Operations,ICDM 2014, LNAI
8557, pp. 214–227, 2014.© Springer International
Publishing Switzerland 2014
[8] A Survey on Machine Learning: Concept,Algorithms
and Applications,Kajaree Das1Rabi Narayan
BeheraInternational Journal of Innovative Research in
Computerand Communication Engineering,Vol. 5,
Issue 2, February 2017.
[9] Machine Learning Algorithms: A Review AyonDey;
International Journal of Computer Science and
Information Technologies, Vol. 7 (3) , 2016, 1174-1179.
[10] A novel Naive Bayes model: Packaged HiddenNaive
Bayes,YaguangJi,Songnian Yu,Yafeng Zhang; ©2011
IEEE
[11] THE RESEARCH OF THE FAST SVM CLASSIFIER
METHOD;YUJUN YANG 1,2,3, JIANPING LI 1, YIMEI
YANG ©2015 IEEE;
[12] Classification with Learning k-Nearest
Neighbors;JormaLaaksonen and ErkkiOja ,1996 IEEE
[13] A Research Study on Unsupervised Machine Learning
Algorithms for Early Fault Detection in Predictive
Maintenance,NagdevAmruthnath,Tarun Gupta,2018
5th International Conference on Industrial Engineering
and Applications.
[14] Dimension Reduction with Extreme LearningMachine
LiyanaarachchiLekamalageChamaraKasun, Yan Yang,
Guang-Bin Huang and Zhengyou Zhang .IEEE
[15] The Role of Fuzzy Logic in Decision Making
Process;KatarínaValášková1,a,*,TomášKlieštik2,band
Mária Mišánková;2nd International Conference on
Management Innovation and Business Innovation.
[16] Fuzzy Based Machine Learning: A Promising
Approach; Sujamol S., Sreeja Ashok and U Krishna

1125
IJSTR©2019
www.ijstr.org