Enhancing QoS of Network Traffic Based on 5G Wireless Networking Using Machine Learning Approaches

Abstract

5G wireless networks are based on heterogeneous networks. Heterogeneous networks offer a higher quality of service (QoS) and let you better utilize the resources of the network. Control of traffic on a network is complicated when a multiplicity of heterogeneous networks are present. When different protocols and data transmission rates are used, heterogeneous networks face the problem of managing and managing network traffic appropriately. In this paper our objective is to reduce Network Traffic and improve QoS for 5G wireless Network thus we have discuss some Supervised and Unsupervised Algorithm of Machine Learning Approach. So we have implemented K-mean algorithm in this paper will reduce traffic and improve the efficiency of 5G wireless Network. The K-mean is an iterative grouping technique that moves data objects between cluster sets until one desired set is reached. The dataset for K-mean Algorithm divides the traffic into two classes and then weighted mean is calculated for each cluster until the resultant output is identical weighted mean. If there are two clusters have identical weighted mean then there are no changes in cluster of classes.KeywordsHeterogeneous networkTraffic classificationMachine learningSupervised learningUnsupervised learning

Full text

Ritika Mehra
Phayung Meesad
Sateesh K. Peddoju
Dhajvir S. Rai (Eds.)
First International Conference, ICCISC 2022
Dehradun, India, June 10–11, 2022
Revised Selected Papers
Computational
Intelligence and
Smart Communication
Communications in Computer and Information Science 1672

Communications
in Computer and Information Science 1672
Editorial Board Members
Joaquim Filipe
Polytechnic Institute of Setúbal, Setúbal, Portugal
Ashish Ghosh
Indian Statistical Institute, Kolkata, India
Raquel Oliveira Prates
Federal University of Minas Gerais (UFMG), Belo Horizonte, Brazil
Lizhu Zhou
Tsinghua University, Beijing, China

More information about this series at https://link.springer.com/bookseries/7899

Ritika Mehra ·Phayung Meesad ·
Sateesh K. Peddoju ·Dhajvir S. Rai (Eds.)
Computational
Intelligence and
Smart Communication
First International Conference, ICCISC 2022
Dehradun, India, June 10–11, 2022
Revised Selected Papers

Editors
Ritika Mehra
Dev Bhoomi Uttarakhand University
Dehradun, India
Sateesh K. Peddoju
Indian Institute of Technology Roorkee
Roorkee, India
Phayung Meesad
King Mongkut University of Technology
Bangkok, Thailand
Dhajvir S. Rai
College of Engineering Roorkee
Roorkee, India
ISSN 1865-0929 ISSN 1865-0937 (electronic)
Communications in Computer and Information Science
ISBN 978-3-031-22914-5 ISBN 978-3-031-22915-2 (eBook)
https://doi.org/10.1007/978-3-031-22915-2
© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2022
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, expressed or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface
A two-day International Conference on Computational Intelligence and Smart Commu-
nication (ICCISC 2022) was organized by the School of Computer Science and Engi-
neering at Dev Bhoomi Uttarakhand University (DBUU), Dehradun, during June 10–
11, 2022. This conference was organized in association with Springer. Our professional
partners for this conference were the ACM Jaipur Chapter and the Dev Bhoomi Uttarak-
hand University CSI Student Chapter, and it was sponsored by the Uttarakhand State
Council for Science & Technology and the Uttarakhand Science Education & Research
Centre, Dehradun, India.
The aim of the conference was to provide a platform for researchers and practitioners
from both academia and industry to meet and share cutting-edge developments in the field
of computational intelligence and smart communication. It also focused on all aspects
of computation intelligence and data sciences with modern and emerging computational
topics.
ICCISC 2022 provided an excellent international forum to share knowledge as well
as their findings in theory, methodology, and/or applications relevant to the confer-
ence themes. The conference featured paper presentations in addition to the keynote
addresses from prominent speakers on related state-of-the-art technologies. The con-
ference benefited the delegates by helping them to add to or improve their skills and
knowledge. The networking during the conference also laid foundations for possible
future collaborations. ICCISC 2022 provided an invaluable platform to raise awareness
about forthcoming innovations in diverse fields of computational intelligence and smart
communication.
The conference brought together a community of international researchers, industrial
experts, and academicians. It not only was restricted to paper presentations but also paved
the way for subsequent discussions on the latest trends in research and development
linked to the conference themes and allied areas.
The conference featured keynote addresses by prominent people, including experts
from the Cloud Lab at the University of Melbourne, Australia, the Artificial Intelligence
Research Institute (IIIA-CSIC), Spain, the Multimedia Data Analytics and Processing
Research Unit at Chulalongkorn University, Thailand, and the AI and Cybersecurity
Research Centre at Staffordshire University, UK.
The main conference themes were as follows:
•Track 1: Wireless Sensor Networks and Computing Technologies
•Track 2: Networks, Security and Privacy
•Track 3: Smart Communication and Technology
•Track 4: Emerging Computing
Topics of interest included the following subthemes: block chain, deep learning,
pattern recognition, modeling and simulation, natural language processing, internet of

vi Preface
things, soft computing, artificial intelligence, quantum computing, cloud computing, fog
computing, cyber security, sentiment analysis, wireless sensor networks, signal process-
ing, intelligent communications and networking, software defined networks, 5G net-
works, mobile and optical broadband, e-health, real time networks, satellite and space
communication, radar and microwaves, secure and energy efficient networks, cognitive
radio and cognitive networks, multimedia communication, intelligent control, robotics,
and smart embedded systems.
The review process is one of the major components that governs the quality of
research being shared and the success of the event as well, thus making it a very critical
part of a conference. To maintain transparency and to ensure that high standards and
ethics of research are followed, the submission of papers was performed through the
EasyChair conference management system. This platform has an included feature for
plagiarism checks, which are performed using the Turnitin plagiarism software tool.
Papers with a plagiarism coefficient of more than 30% plagiarism get rejected automat-
ically. Papers with coefficients below 30% but above 12% were returned to the authors
for revision. Papers having plagiarism coefficients below 12% in the literature review
section (if any) were considered for presentation in conference provided they met other
norms related to plagiarism.
A single-blind review process was followed, with each paper assigned to three inde-
pendent reviewers. Most of the reviewers were external to ensure the quality standards
of the conference. The process mandatorily required three reviews to be completed—
no paper was considered for presentation unless all three reviews were received—and
efforts were made to reassign papers in cases where a reviewer declined or expressed
unavailability for the process. After receiving the reviews for a paper, it was judged on
the basis of positive reviews and comments, and wherever applicable minor or major
modifications as suggested by the reviewers were communicated to authors and the
paper(s) revised as necessary.
If two or more reviews were found satisfactory, then only the paper was sent to the
General Chair for further verification, and the acceptance or rejection of the paper was
at the sole discretion of the General Chair, whilst keeping the reviewers’ comments in
mind. The conference received 106 papers from authors for consideration and, after the
stringent review process, only 56 were shortlisted for presentation. Further, only nine
research articles were considered for publication in Springer’s CCIS series. Out of these
nine, eight are full length papers and one is a short paper. We hope that you enjoy reading
the selected papers.
October 2022 Ritika Mehra
Phayung Meesad
Sateesh K. Peddoju
Dhajvir S. Rai

Organization
General Chairs
Sanjay Bansal Dev Bhoomi Uttarakhand University, India
Preety Kothiyal Dev Bhoomi Uttarakhand University, India
Raj Kishore Tripathi Dev Bhoomi Uttarakhand University, India
Program Committee Chairs
Ritika Mehra Dev Bhoomi Uttarakhand University, India
Dhajvir Singh Rai Dev Bhoomi Uttarakhand University, India
Steering Committee
Jean-Paul Van Belle University of Cape Town, South Africa
Bhuvanesh Unhelkar University of South Florida, USA
Ankit Agarwal Northwestern University, USA
Phayung Meesad King Mongkut’s University of Technology,
Thailand
Sateesh K. Peddoju Indian Institute of Technology Roorkee, India
Waralak V. Siricharoen Silpakorn University, Thailand
S. Gomathi UK International Qualifications Limited, UK
R. C. Bansal University of Sharjah, UAE
Michael Pecht Maryland University, USA
Sachin R. Jain Oklahoma State University, SUSA
Ahmed Elngar Beni-Suef University, Egypt
Sanjeevi Padmanaban Aalborg University, Esbjerg, Denmark
Ahmed J. Obaid University of Kufa, Iraq
Balachandra Pattanaik Wollega University, Ethiopia
Ali Musrrat King Faisal University, Saudi Arabia
Mohammad Shoab Shaqra University, Saudi Arabia
Rhonnel S. Paculanan University of Makati, Philippines
Kourosh Ahmadi Auckland Institute of Studies, New Zealand
Rocha Alvaro University of Lisbon, Portugal
Pao Ann Hsuing National Chung Cheng University, Taiwan
Durga Toshniwal Indian Institute of Technology Roorkee, India
Kunwar Vaisla Bipin Tripathi Kumauni Institute of Technology,
India

viii Organization
Vishal Jain Sharda University, India
Manoj Kumar Shukla Harcourt Butler Technical University, India
Aman Jatin Amity University, Gurgaun, India
Sandeep Vijay Tula Institute, India
R. K. Bharti Bipin Tripathi Kumaon Institute of Technology,
India
Ajit Singh Bipin Tripathi Kumaon Institute of Technology,
India
Mayank Aggarwal Gurukul Kangri Vishwavidyalaya, India
Amit Aggarwal Abdul Kalam Institute of Technology, Tanakpur,
India
Pramod Kumar Krishna Engineering College, India
Vishal Kumar Bipin Tripathi Kumauni Institute of Technology,
India
S. C. Sharma Indian Institute of Technology Roorkee, India
T. S. Arora National Institute of Technology, India
Vibhash Yadav REC Banda, India
Umesh Chandra Banda University of Agriculture and Technology,
India
Anish Gupta Academy of Business and Engineering Science,
India
Vaisla Kunwar Bipin Tripathi Kumauni Institute of Technology,
India
Sudhakar Chauhan National Institute of Technology Kurukshetra,
India
Kapil Gupta National Institute of Technology Kurukshetra,
India
Ved Prakash Amity University, Haryana, India
Nilam Choudhary Jaipur Engineering College and Research Centre,
India
Baldev Singh Vivekananda Global University, India
Suresh Kumar Manav Rachna Institute of Research and Studies,
India
Das Nripendra Manipal University Jaipur, India
Vijay Bhaskar Semwal Maulana Azad National Institute of Technology,
India
Prakash S. Bharath Institute of Higher Education and
Research, India
Gaurav Verma Jaypee Institute of Information Technology, India
Surya Prakash Thapar University, India
R. Dhanasekaran Syed Ammal Engineering College, India
M. K. Sharma Amarpali Group of Institutions, India
Dinesh Goel Poornima University, India

Organization ix
Parma Nand Sharda University, India
Sachin Sharma Graphic Era University, India
Program Committee
Diago Galar Lulea University of Technology, Sweden
Omar H. Alhazmi Taibah University, Saudi Arabia
Anand Nayyar Duy Tan University, Vietnam
Pham Quoc Cuong HCMUT-VNUHCM, Vietnam
Felix J. Garcia Clemente University of Murcia, Spain
G. E. Alexender Cristina Universidad Technica Particular de Loja, Ecuador
Sameeka Saini Dev Bhoomi Uttarakhand University, India
Manik Sharma DAV University, India
Jeetendra Pande Uttarakhand Open University, India
Bagwari Ashish WIT Dehradun, India
Gunjan Bhatnagar Dev Bhoomi Uttarakhand University, India
Vivek Arya Gurukul Kangri University, India
Vipul Sharma Gurukul Kangri University, India
Bhawna Parihar Bipin Tripathi Kumaon Institute of Technology,
India
Poonam Chhimwal Bipin Tripathi Kumaon Institute of Technology,
India
Saurav Mishra Dehradun Institute of Technology, India
Banit Negi GBPIET, India
Alka Dikshit Himachal Pradesh University, India
Ashish Nayar IITM, India
Deepak Sharma Jagan Institute of Management Studies Delhi,
India
Jitendra Rauthan GBPIET, India
Ekta Upadhayay Dev Bhoomi Uttarakhand University, India
Varun Uniyal GBPIET, India
Sunil Mankotiya Himachal Pradesh University Shimla, India
Nishta Kapoor Rajeev Gandhi Government Degree College
Shimla, India
Anurag Jain University of Petroleum and Energy Studies, India
Shamik Tiwari University of Petroleum and Energy Studies, India
Pooja Munjal Delhi University, India
Deepesh Rawat SRHU Dehradun, India
K. C. Mishra WIT Dehradun, India
Ajit Rathor Ajay Kumar Garg Engineering College, India
Anuj Sharma Gurukul Kangri University, India
Anupama Mishra Swami Rama Himalayan University, India
D. C. Pandey Graphic Era University, India

x Organization
Vivek Kumar Gupta Dehradun Institute of Technology, India
Vijay Shankar Sharma Manipal University, India
Nemi Chand Barwar Mugneeram Bangur Memorial Engineering
College, India
Additional Reviewers
Mukesh Joshi
Anvesha Katti
Purnendu Bikash Acharjee
Rajeev Kumar
Akhilesh Kumar Sharma
Gaurav Verma
Sandeep Budhani
Kapil Joshi
Gaurav Aggarwal
Anuj Kumar
Shilpa Srivastava
Aruna Pavate
Kanchan Dabre
Vaibhav Ranjan
Anupama Chadha
Ram Narayan
Sunil Kumar
Wiqas Ghai
Amit Kishor
Samya Muhuri
Atul Garg
Abhineet Anand
Mandeep Kaur
Nishant Mathur
Ahmed A. Thabit
Sanjeev Pippal
Madhulika Mittal
Vivek Ary a
Jitendra Saturwar
Avdhesh Kumar Tiwari
Pooja Gupta
Gesu Thakur
Ashish Gupta
Sapana Singh
Bharti Sharma
Sridhar Iyer
Abhilasha Chauhan
Deepesh Rawat
Sateesh Kumar
Apurva Sharma
Anirudh Mangore
Pratap Singh
Ganesh Yadav
Shalini Puri
Bhagyashree Shendkar
Vineet Kumar Salar
Michael Albino
Sohit Agarwal
Naveen Tewari
Gaurav Goel
Ajesh F.
Vivek Sharma
Sonika Singh
Rajiv Kumar
Pronab Adhikari
Jefferson Costales
Rajat Goel
Rakesh Saini
Layla H. Abood
Anupama Chadha
Nipur Singh
Prashant Kumar
Kuntal Chowdhury
Yogesh Chauhan
Piyush Anand
Michael Albino
Sunil Pathak
Abid Hussain
Sanjeev Kumar
Srikanta Kumar Mohapatra
Gajanand Sharma
Gourav Bathla
Hussein Jabar Khadim

Contents
Wireless Sensor Networks and Computing Technologies
Enhancing QoS of Network Traffic Based on 5G Wireless Networking
Using Machine Learning Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Shivani Saini, Sharvan Kumar Garg, Pankaj Pratap Singh, Arif Ali,
and Akhilesh Pandey
Soil Classification and Crop Prediction Using Machine Learning . . . . . . . . . . . . . 16
Yuvraj Jangir, Tushar Goyal, Sumit Kandari, and Arshad Husain
Analysis of the Performance of Data Mining Classification Algorithm
for Diabetes Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Vijaylakshmi Sajwan, Monisha Awasthi, Prakhar Awasthi, Ankur Goel,
Manisha Khanduja, and Anuj Kumar
Networks, Security and Privacy Parallel and Distributed Networks
Prediction of DDoS Attacks Using Machine Learning Algorithms Based
on Classification Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Anupama Mishra and Deepesh Rawat
Role of Internet of Things and Cloud Computing in Education System:
AReview ............................................................ 51
Ajay Krishan Gairola and Vidit Kumar
Smart Communication and Technology
An Effective Image Augmentation Approach for Maize Crop Disease
Recognition and Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
M. Nagaraju, Priyanka Chawla, and Rajeev Tiwari
Implementation of Artificial Intelligence (AI) in Smart Manufacturing:
AStatus Review ....................................................... 73
Akash Sur Choudhury, Tamesh Halder, Arindam Basak,
and Debashish Chakravarty
Emerging Computing Computational Intelligence
Flight Fare Prediction Using Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
K. P. Arjun, Tushar Rawat, Rohan Singh, and N. M. Sreenarayanan

xii Contents
Impact of Work from Home During Covid-19 on the Socio-economic
StatusofIndia ........................................................ 100
Poonam Ojha, Sudhanshu Maurya, and Manish Kumar Ojha
Author Index ......................................................... 115

Wireless Sensor Networks
and Computing Technologies

Enhancing QoS of Network Traffic Based on 5G
Wireless Networking Using Machine Learning
Approaches
Shivani Saini1(B), Sharvan Kumar Garg1, Pankaj Pratap Singh2,ArifAli
2,
and Akhilesh Pandey3
1Subharti Institute of Technology and Engineering, Swami Vivekanand Subharti University,
Meerut, India
Shivanisaini792@gmail.com
2School of Computer Science and Engineering, Dev Bhoomi Uttarakhand University,
Dehradun, India
3Uttaranchal School of Computing Science, Uttaranchal University, Dehradun, India
Abstract. 5G wireless networks are based on heterogeneous networks. Hetero-
geneous networks offer a higher quality of service (QoS) and let you better utilize
the resources of the network. Control of traffic on a network is complicated when
a multiplicity of heterogeneous networks are present. When different protocols
and data transmission rates are used, heterogeneous networks face the problem of
managing and managing network traffic appropriately. In this paper our objective
is to reduce Network Traffic and improve QoS for 5G wireless Network thus we
have discuss some Supervised and Unsupervised Algorithm of Machine Learning
Approach. So we have implemented K-mean algorithm in this paper will reduce
traffic and improve the efficiency of 5G wireless Network. The K-mean is an
iterative grouping technique that moves data objects between cluster sets until
one desired set is reached. The dataset for K-mean Algorithm divides the traffic
into two classes and then weighted mean is calculated for each cluster until the
resultant output is identical weighted mean. If there are two clusters have identical
weighted mean then there are no changes in cluster of classes.
Keywords: Heterogeneous network ·Traffic classification ·Machine learning ·
Supervised learning ·Unsupervised learning
1 Introduction
Now days 5G wireless network run on application that requiring high demand for data
rates. Heterogeneous network(HetNets) that can use different power level for trans-
mission to assure the data traffic and heterogeneous utilizes multiple types of access
nodes, offers low power consumption, spectrum efficiency, energy efficiency, and qual-
ity of service, and offers reduced green house gases. In addition to the conventional”
High power antenna “HPN and HetNets Introduces “Low Power Antenna”. The high
power antenna can sever large the geographical area and low power antenna can sever
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 3–15, 2022.
https://doi.org/10.1007/978-3-031-22915-2_1

4 S. Saini et al.
comparatively small the geographical area. Various layers of cells, for example, femto,
full scale, miniature, pico, transfers, different client gadgets and applications interface
with the heterogeneous organization.
Heterogeneous In wireless telecommunications, network expressions can have a vari-
ety of meanings. It could, for example, refer to a pattern of flawless and always-present
interoperability among a variety of multi-reporting protocols (HetNet). Alternative uses
for the term in homogeneity include describing the spatial division of wireless nodes or
users (also known as spatial distribution in homogeneity) [1](seeFig.1).
Explains the lack of clarity in technical writing and peer-reviewed publications could
result from describing the perception of “heterogeneous networks” without providing
that context. Secondary uncertainty may arise as a result of the fact that the “HetNet”
pattern can be studied from a “geometric” perspective as well as [2].
Fig. 1. Heterogeneous network
2 Traffic Classifications
(See Fig. 2) The traffic classification what does mean and why should we are in a word
it’s all about performance We believe the traffic classification is serious to civilizing
your Ethernet. IP Network presentation and the consumer knowledge that is because
present simply so much bandwidth presented on your network and present a grouping
of traffic like voice and finances transaction application are critical and needs to get
through quickly as soon as possible next year may have For other traffic, such as Internet
browsing through video streaming this traffic may be less latency sensitive variation
and then there all the rest which still needs to get there but can probably wait a bit by
classifying traffic before putting on network you make the best use of bandwidth you
have available for illustration purpose its compare it an airplane that is only one quarter

Enhancing QoS of Network Traffic Based on 5G Wireless 5
fall what worse is that you cannot fit all your passengers in the first class on the airplane
so bring if another fill up the first again and fly another one quarter full and so on its.
In networking the same with available bandwidth and the network traffic if you’re
not clarity the traffic before it gets on to the network then all of two gets treated as a
priority this doesn’t make sense now here what does you take certain traffic and say this
is my first class traffic must get through rapidly then this is my business class traffic
it can get through rapidly but can wait a little first class need to get through first what
remain is my best effort traffic which can wait.
Fig. 2. Traffic classifications
3 Related Works
In (see Fig. 3) its shows Internet Traffic classification is divided into three Approaches
are followings.
3.1 Port Based Approach
In a node, many processes will be running and data which are sent or received must reach
the right Process. Every Process in a node is uniquely indentified using port number.
Suppose in a computer there are five process running and one process is requesting to
the data to the another computer replying and that reply must reach the right process
which send the request and reaching the right process which has sent the request the
right process and reaching the right process which has sent the request is done with help
of port address [3]. So, Port number or simply port we called as the communication
end point. In real Scenarios we have two categories of port number fixed port number
(25.80) and dynamic port number (0–65565).UDP also uses port numbers, even though
it is connection less service [4].

6 S. Saini et al.
Fig. 3. Traffic classifications
3.2 Payload Based Approach
This method determines the package by parsing the package sub headers. The packet
payload is parsed bit at a time to find a stream of bits spanning the signature. Some-
times that flow is decisive. In this case, you can name the set of factors precisely. This
machine works continuously to detect P2P traffic and identify system outages [5]. The
real downside of this system is that security laws prevent administrators from evaluating
the payload. It also requires a lot of flexibility and load preparation for traffic identifi-
cation devices. It scans the entire payload, requiring significant processing power and
capacity limitations [6].
3.3 Statistical Based Classification
The pattern recognition system is dividing into two major modes of operation training
and classification. The role of preprocessing module is to segment the required pattern
from the given background, removenoise and normalization it’s and after other operation,
represented the pattern for further processing. On this approach the category of networks
is base absolutely on association altitude style and Network protocol performance. This
technique relies entirely on discovering and verifying host performance patterns at the
transport layer. The advantage of this classification is that it does not require packet
payload access [7,8].
4 Work Flow of Traffic Classification in Networks
This is how network traffic can be classified into different types of site visitor learning
based on any parameter [9]. First, it captures the traffic of the community and extracts

Enhancing QoS of Network Traffic Based on 5G Wireless 7
the characteristics of the chosen information. It after that trains the system using the data
sampling system and finally runs the algorithm and computes the outcome (see Fig. 4).
Fig. 4. Traffic classification approaches
STEP: 1 Network Traffic Capture
At this stage of data collection one of the most important and critical step in data collec-
tion. This step captures network traffic in real time. There are lots of tools for intercepting
network traffic, such as Wire Shark.
STEP: 2 Feature Extraction Selections
The second step in network traffic analysis is the selection of feature extraction. This
includes features extracted from the data collected in the first stage of traffic analy-
sis, such as packet length, packet duration, and time between packet arrival protocols,
etc. Then use the extract function to train a machine learning class. After training the
model and receiving the data, machine learning validates the data and outputs the results
accordingly. Some machine learning algorithm classifiers are trained during data testing
and training respectively.
STEP: 3 Training Process Sampling
The third step in network traffic analysis is sampling of the learning process. Contains
data sets selected for supervised learning. In supervised learning, data are first labeled
to classify unfamiliar network applications [10].
STEP: 4 Algorithm Implementation
The fourth step in network traffic analysis is the implementation of machine learning
algorithms. Implementation steps involved in applying a machine learning algorithm
or classifier to an instance. For example, the use of supervised, unsupervised and semi-
supervised learning algorithms. This article implements the algorithm of SVM and naive

8 S. Saini et al.
Bayes supervised K-Nearest Neighbor algorithm, the unsupervised learning algorithm
is K-Means, DBSCAN.
STEP: 5 Results and Observation
The third step of network traffic analysis result and observation.After applying machine
learning algorithm gives the classifier result.
5 Machine Learning in 5G Network
ML algorithms using statistical techniques that can be enhanced with experience with
the machine. The new scenarios and features of the 5G network traffic described above
use many calls for existing motion control strategy [6]. To resolve these problems, you
can resolve a solution to work around so that you can create a solution directly to the
machine training model, or you can learn the data without using a subsequent rule set
[11] (see Fig. 5) is will discuss 5G traffic management from the point of view of the ML
algorithm: controlled training, unconditionally educational.
Fig. 5. Machine learning workflow
6 Supervised Learning
Supervised learning is a method which we provide the machine a label dataset or in other
word we can say that we provide the machine with a given data set on which it is trained
to perform a future task. Supervised learning creates a comprehensive model that maps
contribution features to desired outputs. In a number of cases, maps are implemented
as a set of limited models, such because case-based inference or Nearest-Neighbor
algorithms.
Supervised learning applied to network management has been reported to shape
network routing path selection, traffic volume prediction, etc. [12]. To solve the 5G
network traffic problem with supervised learning, you should consider the following
steps:

Enhancing QoS of Network Traffic Based on 5G Wireless 9
Step 1: Decide on the type of string example.
Step 2: Assemble the training set.
Step 3: Determine the input representation of the recognizable function.
Step 4: It determines the structure of the well-read function and the algorithm for learning
it.
6.1 Supervised Learning Algorithm
These are the supervised machine learning algorithms following:
a) Naïve Bayes
b) Support Vector Machine
c) K-Nearest Neighbor
6.2 Naïve Bayes
Naive Bayes Most of the generic Bayesian network models used for machine learning.
Bayes’ theorem is used to manage network traffic and accurately classifies network traffic
using flow features provided as training data for the model. Naive Bayesian learning has
no problems with noisy data and can make more accurate predictions (See Fig. 6).
Fig. 6. Naïve Bayes work flow
Step 1: Read the dataset (Collection of IP address)
Step 2: Correlation Based Feature Selection (Relationship-based element subset selec-
tion is used in studies to find subsets of highlights with high-level explicit relationships
and low-level relationships.)

10 S. Saini et al.
Step 3: Naïve Bayes Learner (learn the naïve bayes model).
Step 4: Naïve Bayes Predictor (Use Naïve Bayes model to predict classes)
Step 5: Classification Result.
6.3 K-Nearest Neighbor
The nearest neighbor rule is an extension of the nearest neighbor rule. Most classes of
these K nearest neighbors are class labels assigned to the new sample. The value chosen
for k is significant. If the value of k is correct, the classification accuracy is better than
using the nearest-neighbor algorithm.
Networks can assign cluster values and use K-Nearest Neighbors to classify traffic.
In the K-nearest neighbor method, K can be any integer greater than one. Calculate the
nearest neighbor group for each new data point to classify.
There are flowing step in K – nearest neighbor.
Steps 1: Get data.
Steps 2: Define K Neighbors.
Steps 3: Calculated the Neighbors Distance.
Steps 4: Assign new instance to Majority of Neighbors.
6.4 Support Vector Machine
Support Vector Machine (SVM) is a Supervised Machine Learning Algorithm gener-
ally used to partition a numerical data set into different classes based on mathematical
properties and characteristics. Classification aims to find constraints (or equivalently
minimize classification errors) between different classes using limits on the maximum
distance from a sample to that limit [13].
Classification is then performed along the hyper plane that separates the two classes.
If you need a model that can accurately determine if a cat is a dog by looking at a strange
cat and dog that also has cat characteristics, you can use the SVM algorithm to create
such a model. The development concerned in the SVM classifier is as follows:
Step1: Past Labeled Data.
Step 2: Model Training.
Step 3: Predication.
Step 4: Output.
In a network, first the network trains on past labeled data so that it can learn different
characteristics of the data, It then tests the new data and after that learns how the algorithm
predicts and classifies new received classes. You need to prepare the classifier first and
in that case cross-validate it through the data validation. To get correct predictions with
the SVM classifier, you want to utilize the SVM kernel purpose and tune the parameters.

Enhancing QoS of Network Traffic Based on 5G Wireless 11
7 Unsupervised Learning Algorithm
This is learning to train an output device to respond to a group of patterns as input. Unsu-
pervised learning is used in self-organizing neural networks. You don’t need a teacher
for this training. In this learning method, Related types of input vectors are grouped
mutually not including by means of tanning data to point out what a representative com-
ponent of every one grouping capacity look like, or which group a component belongs
to. During training, the neural network receives input models and classifies them. When
a new input model is applied, the neural network provides an output response indicating
which class the input model belongs to. If no class exists for the input model, a new
class is created.
The study of network traffic management in 5G networks allows the use of traffic
patterns and probabilistic modeling in traffic conditions. Network planning and config-
uration, network traffic, better network planning and configuration forecasting. Failure
of hands-free algorithms used in networks – K-Mean, DBSCAN.
7.1 K-Mean
The K-mean is an iterative grouping technique that moves data objects between cluster
sets until one desired set is reached.A tall degree of similitude is accomplished between
components of a cluster, whereas a tall degree of disparity between components of
diverse clusters is accomplished at the same time.
a) Algorithm
A K-mean partitioning algorithm that expresses the centered of each cluster as the
average of the features in the cluster.
K=Number of clusters.
D={t1,t2,------- tn}:An data set contain n objects.
Output: A set of K Clusters.
b) Method
1) Arbitrary in D, select K features as initial cluster centroids.
2) Replicate
3) (re) allocate all object ti in the cluster where the object is nearly all related, It is
based on the average value of the objects in the cluster and Update the cluster mean.
Analyze the average of the features for each cluster.
4) Repeated pending there is rejection adjusts.
c) Proposed Model for Reduced Network Traffic
Ts =Traffic state.
K=Number of clusters.
M=weighted Mean.
i. Suppose that we have given the following traffic states to cluster. ( 2Ts1, 4Ts2,
10Ts3, 12Ts4, 3Ts5, 20Ts6, 30TS7,11Ts8, 25Ts9,) and K =2.

12 S. Saini et al.
ii. We initially assign th means to the first two values M1 =2 and M2 =4.
iii. Using Euclidean distance initially K1 ={2Ts1, 3Ts5} and K2 ={4Ts2, 10Ts3,
12Ts4, 20Ts6, 30TS7, 11Ts8, 25Ts9}
iv. The Value 3 is equidistant from both means, so K1 is arbitrarily chosen.
v. Now, means are recalculated to get M1 =2.5, and M2 =16.
vi. Objects are assigned again to the crew clusters having K1 ={2Ts1, 3Ts5, 4Ts2}
and K2 ={10Ts3, 12Ts4, 20Ts6, 30TS7, 11Ts8, 25Ts9,} Continuing this we
obtain the following.
vii. The clusters in the last two steps are identical.
viii. This will yield identical means and thus the means same and no changes in clusters.
This will provide identical and therefore identical means and no variation in clusters
(Figs. 7,8,9and 10) (Table 1).
Tabl e . 1 . Variation in clusters
M1M2K1K2
318 {2Ts1,3Ts
5,4Ts
2, 10Ts3} {20Ts6, 30TS7, 11Ts8, 25Ts9}
4.75 19.6 {2Ts1,3Ts
5,4Ts
2, 10Ts3, 11Ts8, 12Ts4} {20Ts6, 30TS7, 25Ts9}
725 {2Ts1,3Ts
5,4Ts
2, 10Ts3, 11Ts8, 12Ts4} {20Ts6, 30TS7, 25Ts9}
7.2 DBSCAN
Noise-based spatial density clustering is a density-based clustering algorithm that uses
dense functional areas. Parameters used in the DBSCAN algorithm.
Esp, score, minimum DBSCAN clusters reach the density directly and are
formed at the midpoint where the density can be reached. Collect data from online
tools, select features based on the package, and adapt your model to testing and training.
Finally, forecasts and reviews.
Let X ={x1, x2, x3, …, xn} be the set of data points. DBSCAN requires two.
Parameters: (eps) and the smallest number of points needed to.
Form a cluster (Minpts).
1) The algorithm proceeds by randomly selecting a point from the data set (waiting for
all points to have been accessed. Access).
2) If there are at smallest amount ‘minPoint’ points inside the radius ‘ε’Uptothis
point, all these points are considered part of the same cluster
3) Then the clusters are extended by how to recursively iterate the neighborhood
calculating for both neighbor point.

Enhancing QoS of Network Traffic Based on 5G Wireless 13
Fig. 7. Initial data (K =2)
Fig. 8. Phase – 2 (Finding the Neighbors and voting for label)

14 S. Saini et al.
Fig. 9. Phase -3 (Finding the Neighbors and voting for label)
Fig. 10. Phase-4 (Finding the Neighbors and voting for label)

Enhancing QoS of Network Traffic Based on 5G Wireless 15
8 Conclusions
Heterogeneous networks are the foundation of 5G networks when traffic on the network
plays a major role in disrupting network performance. This article described ML for
5G traffic control, including supervised and unsupervised learning. Supervised learning
algorithm. When we use Naive Bayes algorithm its provide less accuracy because If
there is a variable in the test dataset that is not in the tan dataset, the naive Bayes model
assigns it a probability of zero and makes no predictions about it and Support vector
Machine is not appropriate for huge amount of dataset for the reason that of its high
training time and SVM performance was not good in case of overlapping classes. In
unsupervised learning algorithm we using the k-mean algorithm are used reduce the
network traffic using the method Clustering that gives the more accuracy.
References
1. Soldani, D., Manzalini, A.: Horizon 2020 and beyond: on the5G operating system for a true
digital society. IEEE Veh. Technol. Mag. 10(1), 32–42 (2015)
2. Liu, Y., Zhang, Y., Yu, R., Xie, S.: Integrated energy and spectrum harvesting for 5G wireless
communications. IEEE Network 29(3), 75–81 (2015)
3. Shafi, M., et al.: 5G: A tutorial overview of standards, trials, challenges, deployment, and
practice. IEEE JSAC 35(6), 1201–1221 (2017)
4. Dhote, Y., Agrawal, S., Deen, A.J.: A survey on feature selection techniques for internet
traffic classification. In: 2015 International Conference on Computational Intelligence and
Communication Networks (CICN). IEEE (2015)
5. Dzulkifly, S., Giupponi, L., Sai, F., Dohler, M.: Decentralized Q learning for uplink power
control. In: IEEE International Workshop on Computer Aided Modelling and Design of
Communication Links and Networks, pp. 54–58. IEEE (2015)
6. Li, R., et al.: Intelligent 5G: when cellular networks meet artificial intelligence. IEEE Wirel.
Commun. 24, 175–183 (2017)
7. Alnwaimi, G., Vahid, S., Moessner, K.: Dynamic heterogeneous learning games for oppor-
tunistic access in LTE-based macro/femtocell deployments. IEEE Trans. Wirel. Commun.
14(4), 2294–2308 (2015)
8. Challita, U., Dong, L., Saad, W.: Deep learning for proactive resource allocation in LTE-U
networks. In: European Wireless 2017- 23rd European Wireless Conference (2017)
9. Shafiq, M., et al.: Network traffic classification techniques and comparative analysis using
machine learning algorithms. In: 2016 2nd IEEE International Conference on Computer and
Communications (ICCC). IEEE (2016)
10. Mole, P.V.: Towards 5G Enabled Traffic Management Systems: A Literature Review
11. Chih-Lin, I., Han, S., Xu, Z., Wang, S., Sun, Q., Chen, Y.: New paradigm of 5G wireless
internet. IEEE J. Sel. Areas in Commun. 34(3), 474–482 (2016)
12. Fu, Y., et al.: Artificial intelligence to manage network traffic of 5G wireless networks. IEEE
Network 32(6), 58–64 (2018)
13. Chiti, F., Fantacci, R., Giuli, D., Paganelli, F., Rigazzi, G.: Communications protocol design
for 5G vehicular networks. In: Xiang, W., Zheng, K., Shen, X.( (eds.) 5G mobile commu-
nications, pp. 625–649. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-34208-
5_23

Soil Classification and Crop Prediction Using
Machine Learning
Yuvraj Jangir(B), Tushar Goyal, Sumit Kandari, and Arshad Husain
Department of Computer Science, DIT University, Dehradun, Uttarakhand, India
Yuv.rraj786@gmail.com
Abstract. Soil classification is the process in which soil is segregated according
to its physical and chemical properties. This process can be achieved manually
or using a machine learning algorithm. The use of machine learning algorithms
has been on the rise in recent years due to their accuracy. They can classify soils
with more precision than humans can manually, by considering many factors such
as pH, organic matter content, and particle size distribution. We here proposed a
model to classify soil and to predict the most suitable crops using various algo-
rithms of machine learning like Convolutional Neural networks (CNN), Decision
Trees, Naive Bayes. Soil and crop datasets are used, they comprise of different
geographical and physical. Parameters. The module is tested on manually created
datasets and results are obtained.
Keywords: Soil types ·Machine learning ·Convolutional neural network
(CNN) ·Decision tree classifier
1 Introduction
1.1 Objective of Study
To alleviate the agricultural crisis in its current state, it is necessary to put in place
better recommendation systems to alleviate the crisis by helping farmers make informed
decisions before planting begins. The main objective of this experiment was to classify
and predict the suitability of different types of soils on various crops based on predictor
variables like temperature, rainfall, and location. Prediction obtained using decision trees
algorithm. It is important to note that this algorithm works only when a set number of
input values are set up in the form of training samples, which gives a result prediction for
each sample in accordance with that input setting, which can then be compared against
ground truth at other input settings (some samples may not have the rule applicable).
As such, each predictor variable was considered as a feature to define the suitability of
a given map. A decision tree was then constructed to mimic this process and so predict
the result for an input location.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 16–21, 2022.
https://doi.org/10.1007/978-3-031-22915-2_2

Soil Classification and Crop Prediction Using Machine Learning 17
1.2 Related Works
In order to predicts soil type and based on the prediction, suggests suitable a model is
proposed which includes several machine learning algorithms are used for soil classifica-
tion. Experimental results show that the proposed SVM (support vector machines) based
method performs better than many existing methods [1]. Soil dataset and crop dataset
were used to classify the soil. Soil dataset contains class labelled chemical feature of soil.
The crop suggestion dataset contains class labelled crop suggestion attributes [2]. The
soil image has been analyzed using various image preprocessing. Soil color is identified
using statistical properties such as mean of Red, Blue, Green (RGB) values of image
pixels. A classification of soil based on feature extraction of soil color, soil pH values
and texture by using Support Vector Machine classifier. Here, the one-against-all SVM
method is used for classification. Recommended which crops were suitable for tested
soil image based on feature extraction by using image processing [3]. Digital image
processing and Image analysis technology has been used in which suggestions of crops
to be grown in that soil type. The image of the soil/land is clicked using phone camera
and submit it. An Image is a two - dimensional signal. Image processing is a method
to perform some operations on an image, to get an enhanced image or to extract some
useful information [4]. A model has been proposed for predicting the soil type and sug-
gesting a suitable crop that can be cultivated in that soil. The model predicts soil fertility
and 8 other properties of agricultural land [5]. A comprehensive system is described for
soil classification in which different images of soil samples are captured. The features
of each type of soil are collected and are stored in a separate database. This database
is later used in the final stage for soil classification [6]. In order to efficiently classify
the soil instances and maps the soil type to the crop data to get better prediction with
higher accuracies. Soil prediction involves types of crop classifications and geographical
attributes. It also aims at creating a system that processes the real-time soil data to predict
the crops with higher accuracy [7].
2 Proposed Framework
In present work, we propose a novel methodology for soil classification using Convo-
lutional Neural Network (CNN). (Fig. 1.) The generated classifiers were validated with
the accuracy of more than 90%. Secondly, we discussed about crop prediction using
decision tree-based classifier.
In this paper, we described a proposed architecture for soil classification using image
samples. Soil classification is done by analyzing different types of soil image sam-
ples. Enough training samples are needed to classify soil samples. Training samples
should be selected carefully. After selecting the training dataset, all the images are then
passed through Image Processing which includes resizing and augmentation process.
After applying Image Processing on all images, the resultant images are passed through
Convolutional Neural Network (CNN) for classification.
Convolutional Neural Network (CNN) Is a powerful algorithm for image processing.
These algorithms are currently the best algorithms we have for the automated processing
of images. CNN is a directed acyclic graph with four main layers, which are: input layer,

18 Y. Jangir et al.
Fig. 1. Architecture of the system
filter layer (convolutional layer), pooling layer and output layer to provide better accuracy
of classification. CNN is used to classify the soil samples by extracting various features
from the image. (Fig. 2.) The dataset used in the paper contains nearly 200 cropped
images from different soil types. It is a type of machine learning algorithm that allows
for classification and prediction tasks. It is used to classify inputs into several categories
and predict an output based on given inputs. Images contain data of RGB combination.
The computer does not see an image, all it sees is an array of numbers. Color images are
stored in 3-dimensional arrays. The first two dimensions correspond to the height and
width of the image (the number of pixels). The last dimension corresponds to the red,
green, and blue colors present in each pixel.
Fig. 2. Wor kin g o f CNN

Soil Classification and Crop Prediction Using Machine Learning 19
The output of CNN is compared with a threshold value and if it is greater than the
threshold value then it will be classified into class1 otherwise it will be classified into
class2. The classification process is done using Python 3.10.4 and Keras 2.8.0 lucid
framework. The evaluated accuracy of this method is greater than 90%, which means
the method is reliable, accurate and fast. Then, the trained model is stored in h5 format,
which is then used to predict the soil, using the image provided in the input.
In the crop prediction model, we first created a dataset with the following parameters:
states, rainfall, ground water, temperature, soil type, season, and crop, and stored it in
a csv file. After creating the dataset, we split the data for the training data set and the
testing data set separately. After that, we used the decision tree classifier to train the crop
prediction model. The performance of the model was measured by the accuracy value.
Decision Tree Is a supervised prediction method which is widely used because they can
easily obtain input data and are suitable for classifying the data into a wide range of
categories. Decision trees can be used as a type of classifier or regression model that
uses binary trees to predict outcomes. The key advantage of decision trees is that they
can be easily implemented and interpreted. Decision tree models have been shown to
be effective in many real-world applications. Decision tree classifiers are used in many
fields, such as pattern recognition, data mining, machine learning and bioinformatics.
Decision trees can be used to predict categorical outcomes, and they frequently have an
option to do regression functions.
In this study, we presented a real-world dataset that has the names of various crops.
We also had the opportunity to determine two meteorological inputs: rainfall and ground
water level. These two inputs were necessary to help us build a crop prediction model
using decision trees. After training the model, we stored the model in a.sav format, which
is then used to predict the crop, using the parameters provided in the input and the soil
type predicted using soil classification model.
3 Result and Discussion
3.1 Dataset Collection
Multiple datasets are used to train and obtain relevant results. All the datasets used are
custom datasets; built and structured according to the requirements of the algorithm and
the proposed test cases. Below is the list and types of the datasets used-
1. Soils.zip (Soil Image Dataset) - Contains about 150–200 images of different types
of soil which are used for agriculture and found in the Indian subcontinent.
2. Cat_crops.csv - The CSV file mentioned contains data on various parameters
that were considered when training the machine learning model for the crop
recommendation system.

20 Y. Jangir et al.
Fig. 3. Result obtained
3.2 Results
The below Fig. 3. Represents the accuracy of the soil classification model which was
built using CNN (Convolutional Neural Networks).
After several changes and observations, it has been noticed that after improving the
dataset, the accuracy of the model is also improved.
Fig. 4. Accuracy of different algorithm
In crop prediction model, we provided a csv file which consists of the following
parameters: States, Rainfall, Ground Water, Temperature, Soil Type, Season and Crop.
We compared the results with the previous prediction methods. (see Fig. 4.) We used
4 different classifiers to train the crop prediction model, out of which the decision tree

Soil Classification and Crop Prediction Using Machine Learning 21
classifier gave us the best accuracy among them. The accuracies of all the different
classifiers used were:
4 Conclusions and Future Scope
This proposed system is based on an image processing technique where digital images
of the soil samples were processed using convolutional neural network (CNN). In this
study, Decision Tree was used to determine the crop suitability of the soil sample. The
results showed that CNN recommended which crops were suitable for tested image of
the soil samples. So, the proposed method will help farmers to increase the productivity
of yield by identifying suitable crops for the soil samples.
In future perspective, a point location-based rainfall prediction and ground water
detection module can be integrated with the other parameters. This would increase the
overall prediction of suitable crops. Also, the dataset for all these three experiments
consisted of nearly 200 cropped images from different soil types, a larger dataset is
needed. This will increase the accuracy of prediction and classification of soil as well as
crop.
References
1. Rahman, S.A.Z., Mitra, K.C., Islam, S.M.M.: Soil classification using machine learning meth-
ods and crop suggestion based on soil series. In: International Conference of Science and
Technology Computer (2018)
2. Reddy, K.M.A., Chithra, S., Hemashree, H.M., Kurian, T.: Soil classification and crop Yadav,
suggestion. Int. J. Res. Appl. Sci. Eng. Technol. (2020)
3. Yadav, P., Ahire, P.: Soil health analysis for crop suggestions using machine learning.
AEGAEUM J. (2020)
4. Aishwarya, M., Revathy,R., Periasamy, J.K., Srujana, T.: Soil classification and crop suggestion
using machine learning techniques. J. Gujrat Res. Soc. (2019)
5. Saranya, N., Mythili, A.: Classification of soil and crop suggestion using machine learning
techniques. Int. J. Eng. Res. Technol. 9, 671–673 (2020)
6. Chandan, R.T.: An intelligent model for indian soil classification using various machine
learning techniques. Int. J. Comput. Eng. Res. (IJCER) 33, 3005 (2018)
7. Shravani, V., Uday Kiran, S., Yashaswini, J.S., Priyanka, D.: Soil classification and crop
suggestion using machine learning. Int. Res. J. Eng. Technol. (IRJET) (2020)

Analysis of the Performance of Data Mining
Classification Algorithm for Diabetes Prediction
Vijaylakshmi Sajwan1, Monisha Awasthi1(B), Prakhar Awasthi2, Ankur Goel3,
Manisha Khanduja1, and Anuj Kumar4
1Uttaranchal School of Computing Sciences, Uttaranchal University, Dehradun, India
uumonishaawasthi@gmail.com
2Department of Computer Science and Engineering, RIT, Bangaluru, Karnataka, India
3Department of Business Administration, MIET Group, MIT, Meerut, U.P, India
4Uttaranchal Institute of Technology, Uttaranchal Unversity, Dehradun, India
Abstract. The purpose of this paper is to identify solutions for the diagnosis of
diabetes disease by analyzing the patterns found in the data using classification
algorithms such as Decision Tree, SVM, KNN, Naive Bayes, Random Forest,
Neural Network, and Logistic Regression. According to a WHO report, almost
42.2 crores population of the world has diabetes, who are primarily the residents of
low and middle income countries, and diabetes is resulting in around 0.15 crores
of deaths each year globally [1]. To evaluate and discuss the performance of
above-mentioned algorithms, Orange as a data mining tool has been applied. Fur-
thermore, the data set used in this research is the “Pima Indian Diabetic Dataset,”
which is obtained from the University of California, Irvine (UCI) Repository of
Machine Learning datasets. As this study utilized several classifiers to simulate
actual diabetes diagnosis for local and systemic therapy, the results indicated that
Logistic Regression outperforms all other classifiers. The experimental data also
demonstrated the significance of the suggested model in the study. The disease has
been ranked as the fifth-deadliest in the United States, and there is currently no
cure in sight. With the advancement of information technology and its continued
penetration into the medical and healthcare sectors, diabetes cases and symptoms
have become well documented and discussed. The research is original and adds
value to the current studies in the same domain as researchers develop a more
rapid and efficient method of diagnosing the disease, allowing for more timely
treatment of patients.
Keywords: Accuracy ·Diabetes ·KNN ·Logistic regression ·Naive bayes ·
Neural network ·Random forest ·Support vector machine
1 Introduction
Databases are densely packed with hidden data and are designed to aid in intellectual
decision making. Different types of data analysis, such as classification and prediction,
are used to make predictions about future data and to describe the data classes. The
classification is a process that predicts the labels for categorical classes. The labels for
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 22–36, 2022.
https://doi.org/10.1007/978-3-031-22915-2_3

Analysis of the Performance of Data Mining Classification Algorithm 23
this class may be discrete or nominal in nature. Classification techniques classify data
using a training set and class labels [2]. With the rising prevalence of implementations
of various classification and prediction algorithms, there is a need for a central hub
that could evaluate the performance of all classification algorithms as well as provide
information on which classifier is the best [3].
The objective here is to examine various algorithms of machine learning for classifi-
cation using the diabetes data set. ORANGE is also used for this purpose. The purpose of
this paper is to compare ORANGE classifiers on a diabetes dataset. Such techniques are
compared using the results of their ORANGE calculations. We have used the Diabetes
dataset because it is a chronic and one of the dramatically increasing metabolic diseases
in the world. Diabetes mellitus, more generally referred to as diabetes, is a collection of
illnesses (metabolic) characterized by persistently increased levels of sugar in a blood
(beyond a certain limit) and caused by lowering the secretion of insulin or biological
effects, or both. It is a disorder in which the person’s body is not able to metabolize
food in an adequate manner. Diabetes can wreak havoc on a variety of tissues, most
notably the eyes, kidneys, heart, blood vessels, and nerves, resulting in chronic damage
and dysfunction. Diabetes is primarily classified into two segments (types): T1D – Type
1 Diabetes and T2D - Type 2 Diabetes. Type 1 diabetes typically develops in young aged
people (below 30 years of age), and the general symptoms include thirst and urination
again, as well as elevated levels of sugar in the blood. Only must be treated with insulin
as impossible with other oral drugs. Type 2 diabetes is more prevalent in the younger
than younger aged and senior population, and is frequently related with obesity, hyper-
tension, dyslipidemia, arteriosclerosis, and other disorders [4]. Numerous data mining
classification methods have been developed with the goal of classifying, forecasting,
and diagnosing diabetes. However, no meaningful comparison evaluation of the perfor-
mance of such algorithms has been conducted. There has been no research conducted to
determine which of the existing classifier model scans provides the best prediction for
diabetes. The decision tree, Naive Bayes, Random Forest, KNN (K-Nearest neighbours)
and Support vector machines (SVM) classification methods were utilized in this work
to develop classifier models [5].
2 Related Work
According to Aljumah [6], diabetes is a chronic condition that arises when the body
insulin is ineffectively used or when the pancreas produces insufficient insulin. A promi-
nent hormone, Insulin regulates the levels of blood sugar. Unregulated diabetes results
in a rise of blood sugar, which leads to serious vandalism to various body parts and
systems, like the blood vessels and nerves, over time. According to Health informatics,
it is the study of how to collect, retrieve, communicate, store, and utilize health-related
data, knowledge, and information to the best of one’s ability. Barakat et al. [7] defined
how healthcare providers should handle patient information and how citizens should
participate in their own health care. It is now widely recognized as a necessary and
widespread component of long-term health-care delivery. Machine Learning (ML) is
the fastest-growing area in computer science today. When using machine learning in
diabetes related data for prediction, it’s important to remember that this data isn’t being

24 V. Sajwan et al.
collected to address specific research questions; instead, learning algorithms are being
utilized to analyze biomedical data automatically. Song et al. [8] analyzed multiple
categorization algorithms utilizing characteristics such as thickness of skin, pedigree
of diabetes, glucose level, Body Mass Index, patient age, insulin and blood pressure.
Pradeep and Dr. Naveen compared the machine learning algorithms’ performances in
[9] and measured the accuracy of each algorithm. There were accuracy variations in
terms of techniques utilized, pre-processing and after processing of data. It was noticed
that ‘Pre-processing of data’ had better accuracy and overall performance for prediction
of diabetes. In this study, before preprocessing for prediction of diabetes, the Decision
tree algorithm provided better accuracy as compared to other techniques like Random
forest and Support vector machine. According to Loannis et al. [10], Machine learning
techniques, such as the diabetic disorders dataset, have become a significant tool for
predicting diabetes using diverse medical data sets (DD). In this work, SVM, Logis-
tic Regression, and Nave Bayes were used. They used 10-fold cross validation for the
diabetes dataset (DD). The SVM (Support Vector Machine) strategy outperformed the
others in terms of precision and processing, according to the study. For diabetes predic-
tion, Nilashi et al. [11] suggested a CART (classification and Regression Tree) model.
Expectation Maximization (EM) and PCA (Principal Component Analysis) were applied
to pre-process the data and remove noise before applying the rule. The goal of this study
is to design a diabetes decision assistance system. The effect of CART with removal
of noise provided efficiency and enhanced prediction, allowing human life to be saved
from premature demise. A categorization model was suggested by Kamadi et al. in [12].
One of the most typical problems in categorization, they claim, is reduction of data. PCA
(principal component Analysis) was employed in this work for pre-processing of data,
as well as for reduction of data to enhance accuracy. The study employed a modified
DT (Decision tree) and a fuzzy rule to make predictions. They discovered that reducing
the dataset improves the results. Sajida et al. [13] employed the Canadian primary care
sentinel surveillance Network(CPCSSN) dataset and three machine learning models to
detect diabetes at a primary stage in order to save human lives. To predict diabetes,
decision tree (J48), Adaboost, and Bagging were used in this study. Rathore et al. [14]
Diabetic disorder can be detected and predicted. The performance measurements were
examined using R Studio and the Pima Indians diabetes dataset. SVM and Decision Tree
are two machine learning techniques employed. The SVM has an accuracy of 82%.
In [15], S M Hasan Mahmud et al. forecast diabetes. To discover the performance
measurements of the classification algorithms, 10-fold cross validation procedures were
used. The study found that Naive Bayes outperformed the other classifiers, with an
F1 score of 0.74. On the PIMA dataset, Ahuja et al. [16] conducted a comparison
examination of various machine learning techniques, including NB, DT, and MLP, for
diabetic categorization and found MLP to be superior to other classifiers. Fine-tuning
and efficient feature engineering, according to the authors, can improve MLP’s perfor-
mance. Garca-Ordás, M.T. et al. [17] employ min-max normalization and a variant auto
encoder sparse auto encoder to solve data standardization, feature augmentation and
imbalance. MLP was then used for classification, with an accuracy of 92.31%. Without
preprocessing, Bukhari, M.M. et al. [18] state that their ABP-SCGNN (Artificial Back
Propagation Scaled Conjugate Gradient Neural Network) obtained 93% accuracy. [19]

Analysis of the Performance of Data Mining Classification Algorithm 25
is another example of good performance utilizing NN-based models. They looked at
median value imputation (MVI), KNN and an iterative imputer for imputation of the
missing value. Then, to attain an F1-score of 98%, MLP was employed for classifi-
cation. Khanam and Foo [20] employed MVI and Pearson Correlation for selection of
features and missing value imputation. To further standardize the data and eliminate out-
liers, interquartile ranges were used. The classification model based on DNN achieved
an accuracy of 88.6% using several hidden layers. Overall, missing value imputation
and feature selection regarding data pretreatment techniques were seen to be highly
appropriate for prediction of diabetes classification performance. The majority of data
preparation approaches, on the other hand, have been found to perform well when data is
normally distributed. Nonlinear approaches will be better adapted to the problem if the
data does not conform to normalcy assumptions, and they are likely to add significantly
to a classifier’s performance. As a result, this study will look at nonlinear preprocessing
approaches and classifiers for data preprocessing.
3 Methodology
This section describes the classification model’s approach as well as its efficacy in DM
classification. Figure 1summarises the process.
Fig. 1. Methodology of proposed work

26 V. Sajwan et al.
For all of the algorithms (Naive Bayes, KNN,ANN, Logistic Regression,Decision
Tree, Random Forest, SVM), the ‘Confusion Matrix (CM)’ encapsulates the varioussteps
from raw data to grading, data reduction, pre-processing, scoring, and testing.These steps
are described in greater detail in the following subsections as:
A- It describes the data mining toolkit.
B- It describes the database and its attributes.
C- It provides insights into the pre-processing steps.
D- It discusses the process of classification using the algorithms of seven classifications.
3.1 Data Mining Toolkit
To imitate excellent classification techniques, the Orange Data Mining suite of tools
[21] is utilized. Orange was developed as an Open Source Machine Learning (OSML)
framework having in-built visualization of data and analytic capabilities at the University
of Ljubljana’s Bioinformatics Lab. Orange provides a data preprocessing, classification,
regression, clustering, visualization and assessment environment with association rules.
3.2 Collection of Database
The National Institute of Diabetes and Digestive and Kidney Diseases (NIDDKD)
obtained the Pima Indians Diabetes dataset (PIDD) of patients. We would like to express
our gratitude to Vincent Sigillito for supplying the data and short detailing is provided
in Table 1which shows the class distribution in PIDD.
Tabl e 1 . Distribution of classes in the Pima Indians diabetes dataset
Class value Number of instances Relabeled value
0500 Tested_negative
1268 Tested_positive
The NIDDKD owns the PIDD downloaded from Kaggle [22]. Diabetes mellitus may
be identified with the use of this dataset. It has a total of 2000 records, each with eight
characteristics and the class label (outcome). The data set’s description, including its
properties, statistical analysis, and values, are included in Table 2.
These eight characteristics are symptoms that people may or may not have that
indicate their likelihood of having diabetes mellitus.
3.3 Data Preprocessing
Pre-processing is essential to improving model prediction performance. The Orange
toolbox supports a variety of pre-processing techniques [23]. Three different types of
pre-processing approaches are used in this article to increase the dataset’s quality and
eventually the classification models performance.

Analysis of the Performance of Data Mining Classification Algorithm 27
Tabl e 2 . Data set description, properties, statistical analysis and values of data
SNo Attribute name Attribute
description
Data type of attribute Range of attribute
1Preg Pregnancy
frequency
N0to17
2Plas Concentration of
Plasma Glucose
N0 to199
3Pres BP (Blood
Pressure) (mm,
hg)
N0to122
4 Skin Thickness of skin
fold
N0to99
5Insulin 2hseruminsulin
(mm U/ml
N0to846
6Mass BMI – Body
Mass Index
N0 to 67.1
7Pedi Function of
Diabetes pedigree
N0.078 to
2.42
8Age Age of Person (in
yrs.)
N21 to 81
9Outcome Class variable …………. Tested positive, tested
negative
N* - Numeric
•Removal of values which are missing
Due to the fact that the utilized dataset had some missing values, Orange toolkit
presents three methods for imputing values which are missing: eliminate such records,
change them with values which are random, or lastly, change such values with the mean
of other accessible values [24]. As a result, this strategy is selected to be utilized to
eliminate missing values from the applied dataset.
•Selection of Relevant feature
It is critical to choose the most relevant elements. This stage assigns a score to
each characteristic based on its association with the designated diabetes class. From
the dataset, eight characteristics were retrieved. ANOVA is a statistical technique. [25]
Once ANOVA was calculated, it was obtained that thickness of skin and BP are the least
important characteristics and would play a little role in the process of classification;
hence, they were deleted from the features vector, resulting in six rather than eight
features. Figure 2. The table below summarizes the results of the ANOVA test on the
characteristics.

28 V. Sajwan et al.
Fig. 2. Result of ANOVA test on the characteristics
•Normalize the Data
Normalization of data can simplify operations and increase computation perfor-
mance. As a result, the data were normalized to a general scale in a range of zero and
one [26]. Scaling by standard deviation (SD) is one of the methods provided in the
Orange toolbox.
3.4 Data Classification
During this step, the diabetes dataset was classified using six different algorithms. The
investigated classifiers were Naive Bayes, KNN, ANN, SVM, Random Forest, Decision
Tree, Logistic Regression and Adaboost. The data set of features in the data base is
separated into two parts as training was 70% and testing was 30% to guarantee that the
classification process is exactly fit.
•Naive Bayes
Naive Bayes is a statistical learning technique that uses a condensed version of the
Bayes rule to determine the posterior distribution of a category given the input attribute
values of an example case. Prior probabilities for groups and attribute values that are
conditional on categories are calculated using training data frequency counts. Naive
Bayes is a straightforward and fast technique for learning that frequently beats more
advanced methods. Bayesian classification is both a supervised learning technique and
a statistical classification technique. It is capable of resolving diagnostic and predictive
issues [27].

Analysis of the Performance of Data Mining Classification Algorithm 29
•KNN
The KNN algorithm [28] is a simple classification approach. The detection of the
nearest K neighbours during the training phase. The distance between objects and the
value of K, the number of closest neighbours, are calculated using a similarity measure.
•ANN
ANN is a supervised learning method [29] that uses a network of layers to represent
input data, one or more non-linear layers called hidden layers, and finally an output layer
that represents the classification category.
•Random Forest
This classifier creates a collection of decision trees [30], which is a random subset
of the training data. The test object’s final class is chosen to be one that aggregates votes
from the various decision trees.
•SVM
SVM models are a type of supervised learning method that may be used for both
classification and regression issues, but is most frequently used for classification prob-
lems. This classifier is a widely used statistical model that is built on a logistic function
applied to a binary dependent variable in the model [31].
•Decision Tree
A decision tree is a tree structure that resembles a flowchart. It is a method for
classification and prediction that uses nodes and inter-nodes to describe the data. The
root and internal nodes are test cases that are used to distinguish instances with varying
characteristics. Internal nodes are generated as a result of attribute testing. The class
variable is denoted by the leaf nodes [32].
•Linear Regression
Logistic regression is a technique for binary classification. The input variables are
expected to be numeric and to have a Gaussian distribution. It is not required for the
last statement to be true in logistic regression. In other words, the method is capable
of producing acceptable results even when the data is not Gaussian. Each input value
is assigned a coefficient, which is then linearly merged into a regression function and
converted using a logistic function [33].
4 Evaluation & Result
In this part, the results of implemented performance measurements are shown using the
Orange toolkit’s pleasant graphical interface.

30 V. Sajwan et al.
4.1 Setup of Experiments with Results
This subpart explains the procedure of sampling used, the parameters of the classification
model, and the CM for every algorithm.
•Method of Sampling
The developed models’ performance is evaluated using a K-fold cross-validation
sampling approach [27]. The whole datasets are cross-validated tenfold in this article
(2000 records). The data were divided into tenfold samples. The classification model is
trained on seven folds, with the remaining fold serving as a testing set. As a result, for
training the model 70% and for testing the model 30% of data records were utilized.
•Decision Tree
The CM of the Tree classifier is demonstrated in Fig. 3. Out of 500 data points, which
are labeled as ‘0’, the correct classification is for 402 records. Out of 268 data points,
which are labeled as ‘1’, the correct classification is for 142 records.
The Confusion matrix illustrates four critical metrics for evaluating the Decision
Tree Classifier model: true positive (TP), true negative (TN), false positive (FP), and
false negative (FN). Where TP =142, TN =402, FP 126 and FN =98.
•SVM
To learn the model, the attribute space is transformed into a new feature space using
a Radial Basis Function (RBF) kernel. The maximum number of iterations authorized
was 100. Figure 4depicts the SVM classifier’s confusion matrix.
Whereas out of 500 data points, which are labeled as ‘0’, the correct classification
is for 401 records and out of 268 data points, which are labeled as ‘1’, the correct
classification is for 152 records. Again, the values of four critical metrics are TP =152,
TN =401, FP =116, and FN =99.
•KNN
Figure 5illustrates the KNN classifier’s confusion matrix. The nearest neighbours’
numbers was set to five in the KNN model, and the usage of Euclidean distance was
done to calculate the distance between two points, with points weighted according to
their distance from the query point.
We can see in Fig. 5, The CM summarizes four critical metrics for evaluating the
KNN Where TP =156, TN =413, FP =112 and FN =87.
•Random Forest
A forest was incorporated here with 10 decision trees. In Fig. 6, the model’s confusion
matrix is depicted. The CM illustrates four critical metrics for evaluating the Decision
Tree Classifier model, where TP =161, TN =425, FP =107, and FN =75.

Analysis of the Performance of Data Mining Classification Algorithm 31
•Naive Bayes
Whereas out of 500 data points, which are labeled as ‘0’, the correct classification
is for 403 records and out of 268 data points, which are labeled as ‘1’, the correct and
successful classification is for 182 records.
The CM illustrates four critical metrics for evaluating the Naive Bayes Classifier
model, where TP =182 TN =403, FP =86 and FN =97.
•Artificial Neural Network
In this model, back-propagation was applied with a multi-layer perceptron (MLP)
approach. Each buried layer had 200 neurons with a Rectified Linear Unit (ReLu) activa-
tion function. The Adam technique was then employed to efficiently optimise stochastic
weights. In Fig. 8, the con-fusion matrix for the neural network model is shown. Whereas
out of 500 data points, which are labeled as ‘0’, the correct classification is for 431 records
and out of 268 data points, which are labeled as ‘1’, the correct classification is for 157
records. The Confusion matrix summarizes four critical metrics for evaluating an ANN
Classifier model as TP =157, TN =431, FP =111, and FN =69.
•Logistic Regression
This model’s regularization is set to ridges (L2), and the cost strength is set to its
default value of one (C =1). The model’s CM is depicted in Fig. 9.
From 500 data points labeled 0, 442 records were successfully identified, while from
268 data points labeled 1, 151 records were correctly classified.
True positive (TP), true negative (TN), false positive (FP), and false negative (FN)
are four significant metrics used to assess Logistic Regression Classifier model (FN).
TP =151, TN =442; FP =117; FN =58 (Fig. 7).
Fig. 3. CM of tree classifier
•Comparison of Performance
The classification methods performance on the dataset of diabetes is examined and
compared. The following sections contain details on performance measurements and
comparisons.

32 V. Sajwan et al.
Fig. 4. CM of SVM
Fig. 5. CM of KNN
Fig. 6. CM of random forest
Fig. 7. CM of Naïve Bayes

Analysis of the Performance of Data Mining Classification Algorithm 33
Fig. 8. CM of artificial neural network
Fig. 9. CM of logistic regression
•Evaluation Measures of Performance
As mentioned before, the CM illustrates four critical metrics for evaluating classifica-
tion models: true negative (TN), true positive (TP), false negative (FN) and false positive
(FP). These metrics are applied to calculate the following measures of performance:
a) Recall b) Precision c) Accuracy D) F1-measure. These performance metrics are
derived by the use of (TP, TN, FP, and FN). The following metrics are used in this study
to examine and evaluate categorization models:
Accuracy =TP +TN
TP +TN +FP +FN (1)
precision =TP
TP +FP (2)
Re call =TP
TP =FN (3)
F1−measure =2×(precision ×recall )
precision +recall (4)
•Classification Model Comparison
The performance of the implemented classifiers is assessed in this subsection
using the aforementioned metrics. Table 3summarizes the performance metrics for
the classifiers used.

34 V. Sajwan et al.
Tabl e 3 . Measures of performance of applied classifiers
Method AUC CA F1 Precision Recall
Tree 67.9% 70.8% 70.04% 70.2% 70.8%
SVM 75.9% 72% 71.8% 71.6% 72.0%
KNN 78.8% 74.1% 73.8% 73.6% 74.1%
Naïve Bayes 82.9% 76.2% 76.3% 76.4% 76.2%
Random Forest 81.1% 76.3% 75.9% 75.8% 76.3%
Neural Network 82.6% 76.6% 76% 76.0% 76.6%
Logistic Regression 82.9% 77.2% 76.4% 76.7% 77.2%
It also compares the accurate performance of all applicable models. It is self-
evident that Logistic Regression surpasses other classifiers with 77.2% accuracy. Logistic
Regression is followed by a Artificial Neural Network model in second place with an
accuracy of 76.6% and Random Forest in third place with a accuracy of 76.3%. Random
forest is followed by the KNN model in fourth place with accuracy of 74.1%. And SVM
got fifth position with accuracy of 72%. Decision tree with the accuracy of 70.8% is
the worst case. Logistic regression outperforms in all performance measures like AUC,
F1-score, Precision and Recall, which can be shown in Table 3.
5 Conclusion
Automatic diabetes detection is a significant real-world medical issue. Early detection
and management of diabetes are critical. This article demonstrates the use of several
classifiers, including Decision Trees, SVM, KNN, Naive Bayes, Random Forest, Neural
Network, and Logistic Regression, to simulate actual diabetes diagnosis for local and
systemic therapy, as well as presenting relevant work in the field and the outcome indi-
cates that Logistic Regression outperforms all other classifiers. The suggested model’s
usefulness is demonstrated by experimental data. The performance of the strategies was
evaluated in relation to the problem of diabetes diagnosis. Experiments validate the
given model. In the future, it is planned to compile data from several locations across
India and develop a more precise and broad predictive model for diabetes diagnosis.
Future research will similarly focus on accumulating data from a later time period and
identifying additional possible prognostic factors to integrate. The technique might be
expanded and refined to automate the analysis of diabetes.
References
1. https://www.who.int/news-room/fact-sheets/detail/diabetes
2. Amin, D.M., Garg, A.: Performance analysis of data mining algorithms. J. Comput. Theor.
Nanosci. 16(9), 3849–3853 (2019). https://doi.org/10.1166/jctn.2019.8260

Analysis of the Performance of Data Mining Classification Algorithm 35
3. Saichanma, S., Chulsomlee, S., Thangrua, N., Pongsuchart, P., Sanmun, D.: The observation
report of red blood cell morphology in Thailand teenager by using data mining technique.
Adv. Hematol. 2014, 1–5 (2014). https://doi.org/10.1155/2014/493706
4. Canlas, R.D. (2009). Data Mining in Healthcare: Current applications & Issues, Unpublished
Master Thesis, 1–10
5. Iyer, A., Jeyalatha, S., Sumbaly, R.: Diagnosis of diabetes using classification mining tech-
niques. Int. J. Data Min. Knowl. Manag. Process 5(1), 01–14 (2015). https://doi.org/10.5121/
ijdkp.2015.5101
6. Aljumah, A.A., Ahamad, M.G., Siddiqui, M.K.: Application of data mining: diabetes health
care in young and old patients. Journal of King Saud University - Computer and Information
Sciences 25(2), 127–136 (2013). https://doi.org/10.1016/j.jksuci.2012.10.003
7. Barakat, N., Bradley, A.P., Barakat, M.N.H.: Intelligible support vector machines for diagnosis
of diabetes mellitus. IEEE Trans. Inf. Technol. Biomed. 14(4), 1114–1120 (2010). https://
doi.org/10.1109/titb.2009.2039485
8. Komi, M., Li, J., Zhai, Y., Zhang, X.: Application of data mining methods in diabetes
prediction. In: 2nd International Conference on Image, Vision and Computing (ICIVC),
pp. 1006–1010 (2017)
9. Pradeep, K.R., Naveen, N.C.: Predictive analysis of diabetes using J48 algorithm of clas-
sification techniques. In: 2nd International Conference on Contemporary Computing and
Informatics (IC3I), pp. 347–352 (2016)
10. Kavakiotis, I., Tsave, O., Salifoglou, A., Maglaveras, N., Vlahavas, I., Chouvarda, I.: Machine
learning and data mining methods in diabetes research. Comput. Struct. Biotechnol. J. 15,
104–116 (2017). https://doi.org/10.1016/j.csbj.2016.12.005
11. Nilashi, M., Ibrahim, O.B., Ahmadi, H., Shahmoradi, L.: An analytical method for diseases
prediction using machine learning techniques. Comput. Chem. Eng. 106, 212–223 (2017).
https://doi.org/10.1016/j.compchemeng.2017.06.011
12. Kamadi, V.V., Allam, A.R., Thummala, S.M.: A computational intelligence technique for
the effective diagnosis of diabetic patients using principal component analysis (PCA) and
modified fuzzy SLIQ decision tree approach. Appl. Soft Comput. 49, 137–145 (2016). https://
doi.org/10.1016/j.asoc.2016.05.010
13. Perveen, S., Shahbaz, M., Guergachi, A., Keshavjee, K.: Performance analysis of data mining
classification techniques to predict diabetes. Procedia Comput. Sci. 82, 115–121 (2016).
https://doi.org/10.1016/j.procs.2016.04.016
14. Rathore, A., Chauhan, S., Gujral, S.: Detecting and predicting diabetes using supervised
learning: an approach towards better healthcare for women. Int. J. Adv. Res. Comput. Sci.
8(5), 1192–1195 (2017)
15. Mahmud, S.M.H., et al.: Machine Learning Based Unified Frameworkfor Diabetes Prediction.
Association for Computing Machinery. China (2018). https://doi.org/10.1145/3297730.329
7737
16. Ahuja, R., Sharma, S.C., Ali, M.: A diabetic disease prediction model based on classification
algorithms. Annals of Emerging Technologies in Computing 3(3), 44–52 (2019). https://doi.
org/10.33166/aetic.2019.03.005
17. García-Ordás, M.T., Benavides, C., Benítez-Andrades, J.A., Alaiz-Moretón, H., García-
Rodríguez, I.: Diabetes detection using deep learning techniques with oversampling and
feature augmentation. Comput. Methods Programs Biomed. 202, 105968 (2021). https://doi.
org/10.1016/j.cmpb.2021.105968
18. Bukhari, M.M., Alkhamees, B.F., Hussain, S., Gumaei, A., Assiri, A., Ullah, S.S.: An
improved artificial neural network model for effective diabetes prediction. Complexity 2021,
1–10 (2021). https://doi.org/10.1155/2021/5525271

36 V. Sajwan et al.
19. Roy, K., et al.: An enhanced machine learning framework for type 2 diabetes classification
using imbalanced data with missing values. Complexity 2021, 1–21 (2021). https://doi.org/
10.1155/2021/9953314
20. Khanam, J.J., Foo, S.Y.: A comparison of machine learning algorithms for diabetes prediction.
ICT Express 7(4), 432–439 (2021). https://doi.org/10.1016/j.icte.2021.02.004
21. Orange – Data Mining Fruitful & Fun. https://orange.biolab.si/
22. Diabetes –dataset. https://www.kaggle.com/datasets/uciml/pima-indians-diabetes-database/.
Accessed 01 April 2022
23. Pattnaik, P.K., Rautaray, S.S., Das, H., Nayak, J.: Progress in computing, analytics and
networking. In: Proceedings of ICCAN 2017 (2018)
24. Garcia, S., Luengo, J., Herra, F.: Data Preprocessing in Data Mining. Springer (2015). https://
doi.org/10.1007/978-3-319-10247-4
25. Alsalamah, M., Amin, S., Palade, V.: Clinical practice for diagnostic causes for obstructive
sleep apnea using artificial intelligent neural networks. In: Miraz, M.H., Excell, P., Ware, A.,
Soomro, S., Ali, M. (eds.) iCETiC 2018. LNICSSITE, vol. 200, pp. 259–272. Springer, Cham
(2018). https://doi.org/10.1007/978-3-319-95450-9_22
26. Witten, I.H., Frank, E., Hall, M.A., Pal, C.J.: Data Mining: Practical Machine Learning Tools
and Techniques. Morgan Kaufmann (2016)
27. Rennie, J.D., et al.: Tackling the poor asumptions of naive bayes text classifiers. In: Proceed-
ings of the 20th International Conference on Machine Learning (ICML 2003), pp. 616–623
(2003)
28. Chen, G.H., Shah, D.: Explaining the success of nearest neighbor methods in prediction.
Found. Trends® Mach. Learn. 10(5–6), 337–588 (2018). https://doi.org/10.1561/2200000064
29. van Gerven, M., Bohte, S.: Artificial neural networks as models of neural information
processing. Front. Comput. Neurosci. 11, 114 (2017). https://doi.org/10.3389/fncom.2017.
00114
30. Davies, A., Ghahramani, Z.: The random forest kernal and other kiernals for big data from
random partitions (2014). arXiv.1402.4293
31. Smola, A.J., Schölkopf, B.: A tutorial on support vector regression. Stat. Comput. 14(3),
199–222 (2004). https://doi.org/10.1023/b:stco.0000035301.49549.88
32. Rokach, L.: Data Mining with Decision Trees: Theory and Application, vol. 81. World
Scientific (2014)
33. Weisberg, S.: Applied Linear Regression, 4th ed. Wiley (2013)

Networks, Security and Privacy Parallel
and Distributed Networks

Prediction of DDoS Attacks Using Machine
Learning Algorithms Based on Classification
Technique
Anupama Mishra1(B)and Deepesh Rawat2
1Computer Science and Engineering, Himalayan School of Science and Technology, Swami
Rama Himalayan University, Dehradun, India
anupamamishra@srhu.edu.in
2Electronics & Communication Engineering, Himalayan School of Science and Technology,
Swami Rama Himalayan University, Dehradun, India
Abstract. Distributed denial of service attacks often know as network threat is a
severe threat, and are a type of cyber-attack that are directed at a particular system
or network in an effort to make that system or net-work out of reach and unusable
for a period of time. The improved detection of a wide variety of dis-tributeddenial-
of-service (DDoS) cyber threats by utilizing advanced algorithms and a higher
level of accuracy while maintaining a manageable level of computational cost has
consequently emerged as the utmost essential part of detecting DDoS in today’s
world. The DDoS attack that has been launched against the targeted network or
system must be determined in view of defending the machines in a net-work that
has been targeted. In this paper, a number of ensemble classification techniques
are dis-cussed, which combine the performance of various algorithms to improve
overall performance. Using many performance metrics such as a receiver operating
characteristics (ROC) curves, precision, accuracy, recall and F1 scores, we present
and analyzed the performance of algorithms used in our proposed approach.
Keywords: Distributed denial of service attack ·Machine learning ·Random
forest ·Naïve Bayes ·Decision tree
1 Introduction
A denial-of-service attack, according to the World Wide Web security question, is one
that is “de-signed to prevent a computer or network from providing normal services [1].“
The rapid expansion of Internet has resulted the specific type of DoS attack development
that has proven to be extremely effective and difficult to defend against the distributed
DoS attack. This attacks do not originate from a single source, but rather from a number
of spoofed sources that use a variety of attack types in a coordinated effort. Fake Internet
Protocol (IP) addresses, as opposed to real ones, are used to identify computers that are
either unwitting accomplices or that the attacker has control over. Attackers are able to
coordinate deadly attacks on multiple targets at the same time using their own resources
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 39–50, 2022.
https://doi.org/10.1007/978-3-031-22915-2_4

40 A. Mishra and D. Rawat
and the resources of their “zombies,” resulting in greater damage in a shorter amount of
time than they could have done otherwise [2].
A distributed denial of service which is also a cyberattack can bring websites, servers,
and other online services to a crawl. The perpetrator uses multiple computers and devices
to send fraudulent requests to a server, making it appear that the server is being attacked
by a large number of people. The term is some-times used interchangeably with the
term “denial of service attack”, but “DDOS” refers specifically to an attack that uses
multiple sources to flood a target with requests. Some DDOS attacks involve the use
of botnets, which are networks of compromised computers and devices that have been
malware-infected without the users’ knowledge [3].
DDoS is a form of cyberattack where a network is attacked with heavy traffic that’s
create a problem for users to access a website or service. This type of attack is often
used as a tactic to make a target site or service appear overwhelmed or unreliable to
users. DDoS attacks can also be used to force a website or service to a user to a specific
location, where it is under the control of the attacker. DDoS attacks are often used for
online harassment, for example, when a website is under constant attack and can’t im-
prove its performance or stay online [4]. DDoS is a method of disrupting a system in a
network by sending a large amount of data to a server or system from multiple different
sources. The result is that the system or network is not able to handle the load and
ultimately crashes. This makes the system or net-work unavailable to its intended use
[5].
DDoS is a form of cyberattack where multiple hosts are used to bombard a web
server or other net-work target with data, often using a botnet or other network of
malware-infected computers, until the target’s resources are consumed and it is rendered
inaccessible to legitimate users. Unlike a traditional DoS attack, where a single host is
used to flood a target’s resources and crash their services, a DDoS at-tack is much more
complicated and powerful. DDoS attacks can be extremely difficult to stop due to their
decentralized nature. DDoS attacks can be carried out with a handful of hosts or even a
single host, making them much more difficult to detect and investigate [6]. A denial of
service refers to a situation in which a service or resource is utilized to the point of being
rendered unusable or inaccessible to other users. These are often seen in the context of
online gaming servers, but can also affect online banking or shopping services.
A distributed denial of service occurs when a single device or user is able to bring
down a server or network resource. This can be accomplished by distributing a request
across a network, such as when a single computer is used to send an entire file to a website.
Figure 1depicts how this attack [3] had a significant impact on the telecommunications
industry [7]. Current and former employees of several tech companies are accusing
Amazon, Facebook, Google and Microsoft of failing to protect their employees from the
COVID-19 pandemic, with some accusing the companies of endangering their workers’
safety. We’ve seen this story before. In the wake of a string of high-profile layoffs,
employees have accused the companies of not doing enough to protect their workers’
health. But for companies as large as Amazon and Facebook, the risks of a pandemic
are remote. Therefore, the developed technology helps in this field to defend our system
and network from the threat. Machine learning is one of the technologies which is being
used for defensive mechanisms in many applications.

Prediction of DDoS Attacks Using Machine Learning Algorithms 41
Following are the sections that provides an outline for the paper: Sect. 2presents the
related work based on existing defensive mechanisms, our proposed work is discussed
in Sect. 3. Section 4evaluated the research work, and Sect. 5brings the research to its
conclusion.
Telecommunicat
ions
51%
Gaming /
Gambling
23%
Informaon
Technology
and Services
21%
Computer
Soware
3%
Business
Services
2%
Computer
Hardware
0%
Cryptocurrency
0%
Retail
0%
pct_a_bytes
Fig. 1. Impact of distributed denial of service attacks
2 Related Work
Machine learning allows to compare models with different features, which can help to
choose the one that is best for the use of applications. Specifically, we can use a machine
learning model to identify which types of data are most predictive of the outcome of
interest, such as cyberattack.
A defensive approach based on low-cost was proposed by the authors in [6,7]and
focused in calculation of the entropy between benign traffic and DDoS attacks. As an
additional suggestion, the authors Intensity reduction strategy for dealing with attacks has
the following characteristics: The following are three advantages of using this method-
ology over other current methods: The first is that it has a high level of detection. In
addition to having a reduces false alarm, also the capability of detecting small changes
in the environment at a rapid pace along with the mitigate approach.
The authors [7] developed a solution for resolving authentication and security chal-
lenges connected to smart vessels in sea transport. An identity-based approach is used
to authenticate the access for smart vessel and devices. But the method is limited to
maritime transportation.
An auction with many attributes was proposed in [8] to mitigate distributed DDoS
attacks on a net-work. A reputation-based detection approach was proposed, in which
the minimal utility defines the user’s reputation. A payment plan for normal users and

42 A. Mishra and D. Rawat
another for fraudulent users, as well as an identifying mechanism are proposed in addition
with the identification method. In this method, a greedy re-source for allocation strategy
is used to ensure that resources are distributed appropriately among legitimate users.
Differential payment systems are designed to penalize malevolent users that manipulate
their offers in order to obtain the maximum possible share of limited resources.
The authors [9] describe an approach for detecting distributed denial of service
attacks that makes use of Bayesian game theory. It is assumed that the service provider as
well as legitimate users monitor the network in order to collect probabilistic information
to ensure that another user is acting maliciously on their behalf or not. As a result of
having this probabilistic knowledge, both the service provider and authorized users have
the ability to alter their actions and replies in reaction to harmful activity on the network.
The authors propose a Bayesian pricing [10] and auction approach for obtaining Bayesian
Nash Equilibrium points in a variety of settings in which genuine consumers and service
providers benefit from probabilistic knowledge. This is accomplished by taking into
consideration the aforementioned assumptions and facts. In addition to this, a reputation
evaluation and updating system is offered to determine a user’s dependability based on
factors such as the user’s payment history and the amount of time spent participating in
the platform (Table 1).
Tabl e 1 . Comparative table of existing work.
References Techniques Merits Limitation
[10] Used SDN (Software
Defined Network
Detection Rate
is high
Only work on
Volume based DDoS
[11] Worked on IBE (Identity
Based Encryption)
Detection Rate is high Overhead is high
[12] Worked on IBS (Identity
Based Signature) along with
IBE
Detection rate is moderate Overhead is high
[13]BasedonBoosting
Algorithms
Detection rate is moderate False alarms
3 Proposed Work
3.1 Approach
In our paper, we are primarily concerned with data preprocessing, selection of significant
features [13], machine modelling through a classifier, and then finally prediction on
testing dataset. After performance evaluation on results, the research work is concluded.
The approach includes the following activities [14–19]:
Preparation of information: This phase is concerned with preparation of the data
which is comprised of tasks helps on processing the raw data into a clean dataset.

Prediction of DDoS Attacks Using Machine Learning Algorithms 43
If the raw data is not in a usable state at the time of completion, the type and order
of activities may change, and Some of the unrelatable features may be removed. A
few examples of the responsibilities involved are data cleansing, feature selection, and
data transformation. In our work, Fig. 2depicts the best 15 features by using extra
tree classifiers. The selected features are: ACK Flag Count, Inbound, URG Flag Count,
Destination IP, Source IP, Init_Win_bytes_forward, Timestamp, Flow ID, Pro-tocol, Min
Packet Length, min_seg_size_forward, Destination Port, Max Packet Length, Average
Packet Size, and Packet Length Std.
Modelling: In the modelling phase, modelling techniques are applied to the data. This
is done in order to achieve the best possible performance by adjusting the parameters of
the models in question.
As previously indicated, this step is closely tied to data preparation because modelling
might disclose previously unknown data errors. Depending on the situation, the data
preparation method can result in the employment of several models.
Fig. 2. 15 Best Features are selected by using extra trees classifiers

44 A. Mishra and D. Rawat
3.2 Modelling
In order to compare and contrast the datasets, three different supervised learning clas-
sifiers are selected based on a number of parameters [15–17], including the paramet-
ric models and nonparametric models, applications and use of algorithms have been
discussed and used in previous work.
3.2.1 Random Forest
A machine learning algorithm that is used to classify a dataset into a specific category. It
is a combination of many decision trees. The results of the decision trees are combined
in a way that helps reduce the error rate of the classification. This is similar to how a
forest is grown [20–22].
Random forest is a machine learning technique that groups examples together by
their similarity, rather than grouping them by their distance to the target classification.
This is often referred to as “many small decisions,” as opposed to “one big decision,”
which is how other machine learning techniques work. This means that random forest
will, on average, get more things right than other machine learning techniques. However,
it is also more likely to get things wrong.
A machine learning technique that finds patterns in large numbers of variables. For
example, in a medical diagnosis problem, instead of just looking at a patient’s symptoms
and lab results, a machine learning technique might look at millions of different combi-
nations of symptoms and lab results to find patterns that help make a better diagnosis.
In a financial prediction problem, instead of just looking at a stock’s past performance,
a machine learning technique might look at millions of past stock trans-actions to find
patterns that help predict whether a stock will rise or fall. The same is true for any other
problem: finding the right combination of input variables is critical for making accurate
predictions [23].
The random forest technique is often more effective than traditional decision trees,
because it is more likely to capture non-linear relationships in data.
3.2.2 Decision Tree
Decision trees are a machine learning technique that finds patterns in large amounts of
data. In a traditional decision tree, the data is split into two groups: examples that should
be classified as “yes” or “no” in the question being asked, and examples that should be
classified as “yes” or “no” on their own. This is a two-class classification problem. For
example, if the question being asked is “does this dog have fleas?” [24,25].
A machine learning technique that uses decision trees to make a classification. Each
decision tree is built on a subset of the original data. The random forest technique is
often more effective than traditional decision trees, because it is more likely to capture
non-linear relationships in data. Data is split into a number of groups, and each group is
given a separate decision tree.
A machine learning technique that finds patterns in large numbers of variables. For
example, in a medical diagnosis problem, instead of just looking at a patient’s symptoms
and lab results, a machine learning technique might look at millions of different combi-
nations of symptoms and lab results to find pat-terns that help make a better diagnosis.

Prediction of DDoS Attacks Using Machine Learning Algorithms 45
In a financial prediction problem, instead of just looking at a stock’s past performance,
a machine learning technique might look at millions of past stock transactions to find
patterns that help predict whether a stock will rise or fall. The same is true for any other
problem: finding the right combination of input variables is critical for making accurate
predictions. It is a tree-based method that uses the outcome of a single decision tree as the
input for the next tree in the forest. This helps reduce the error rate of the classification.
This is similar to how a forest is grown [26].
Decision trees are one of the most basic machine learning techniques. They are a
machine learning technique that finds patterns in large numbers of input variables. To
build a decision tree, a machine learning technique starts by choosing a subset of the
original data as the root node and then from there, the machine learning technique divides
the original data into subsets, or nodes, based on some criteria, such as variable type or
variable range.
3.2.3 Naïve Bayes
A machine learning technique that uses Bayes theorem to make predictions. The Naive
Bayes ma-chine learning technique assumes that each input variable is independent of
the others. For example, if a machine learning technique is trying to predict whether a
person has a certain disease, Naive Bayes would assume that the presence or absence of
a symptom has no effect on the prediction. This assumption turns out to be surprisingly
accurate in many cases [27,28].
An example of a machine learning technique that uses decision trees is the naïve
Bayes classification machine learning technique and often used in text classification
problems. In a text classification problem, the naïve Bayes machine learning technique
uses decision trees to classify text into different categories. The naïve Bayes machine
learning technique uses decision trees instead of other machine learning techniques
because decision trees are able to capture non-linear relationships in data better than
other machine learning techniques.
Naive Bayes [18,19] used to make predictions. It is often one of the first machine
learning techniques that people learn because it is easy to understand. The Naive Bayes
machine learning technique is also often used as a simple baseline to compare the
accuracy of other machine learning techniques. For ex-ample, if a machine learning
technique is twice as accurate as the Naive Bayes machine learning technique, then it is
likely that the first machine learning technique is a good one.
Naive Bayes is a machine learning technique that is used to make a prediction. It
is a classification technique. For example, if the question being asked is “Will it rain
today? a Naive Bayes technique might be used to predict whether it will rain today. It is
a machine learning technique that finds patterns in large numbers of variables. It works
on bayes theorem by combining the probability that a certain in-put will lead to a certain
output with the probability that a different input will lead to the same output.
4 Result Analysis
We have the following metrics for evaluating the performance of classification machine
learning [29–31]. Performance metrics for machine learning such as precision, recall,

46 A. Mishra and D. Rawat
and f1-score are used to evaluate the quality of the model. These metrics can be used
to compare the performance of different models and to evaluate the impact of different
training regimes on model performance. Precision is a measure of the number of correctly
classified examples. It is calculated as the ratio of the number of examples correctly
classified by the model to the total number of examples in the training set.
The precision is the percentage of the time that an output was actually produced. The
recall is the per-centage of the time that a specific output was actually identified. Table
2,Fig.3and 4are used to show the results as precision, recall , f1 score and accuracy.
Tabl e 2 . Classification report of applied algorithms.
Decision Tree Random Forest Naïve Bayes
Precision Recall F1
Score
Precision Recall F1
Score
Precision Recall F1
Score
Benign 94 98 96 94 100 97 84 100 94
DDoS 100 100 100 100 100 100 100 99 100
75
80
85
90
95
100
105
Precision Recall F1 Score Precision Recall F1 Score Precision Recall F1 Score
Decision Tree Random Forest Naïve Bayes
Performance Matrics of Applied Classifiers
Benign DDoS
Fig. 3. Performance metrics of applied classifiers

Prediction of DDoS Attacks Using Machine Learning Algorithms 47
99.2
99.3
99.4
99.5
99.6
99.7
99.8
99.9
Random Forest Decision Tree Naïve Bayes
Accuracy
Fig. 4. Accuracy of applied classifiers
As shown in Figs. 5,6and 7, the results are discussed and analyzed. All three of
them performed admirably when it came to classifying DDoS traffic. The results of the
performance metric evaluations can be confirmed through the examination of the RoC
Curve.
Fig. 5. RoC curve of Random Forest

48 A. Mishra and D. Rawat
Fig. 6. RoC curve of Naïve Bayes
Fig. 7. RoC curve of Decision Tree
5 Conclusion
In this study, the datasets were classified into binary classification using machine learning
classifiers, and each class was detected and validated properly. A comprehensive analysis
of multiple machine learning algorithms was carried out for the purpose of identifying
DDoS cyber threats, with the Random Forest with the highest accuracy score of 99.80
percent. The naive bayes method achieved 99.42 percent accuracy, while the decision
tree achieved 99.75 percent accuracy in achieving the target. For future work, types of
DDoS attacks can be targeted for classification and prediction in the future.
References
1. Badve, O.P., et al.: Taxonomy of DoS and DDoS attacks and desirable defense mechanism in
a cloud computing environment. Neural Comput. Appl. 28(12), 3655–3682 (2017)
2. Gupta, B.B., et al.: A comprehensive survey on DDoS attacks and recent defense mechanisms.
In: Handbook of Research on Intrusion Detection Systems, pp. 186–218. IGI Global (2020)
3. https://radar.cloudflare.com/notebooks/ddos-2022-q1. Accessed 2 Apr 2022

Prediction of DDoS Attacks Using Machine Learning Algorithms 49
4. Mishra, A., et al.: Security threats and recent countermeasures in cloud computing. Modern
Principles, Practices, and Algorithms for Cloud Security, pp. 145–161. IGI Global (2020)
5. Mishra, A., Gupta, N.: Analysis of Cloud Computing Vulnerability against DDoS. In: Interna-
tional Conference on Innovative Sustainable Computational Technologies (CISCT), pp. 1–6.
IEEE (2019)
6. Mishra, A., Gupta, N., Gupta, B.B.: Defense mechanisms against DDoS attack based on
entropy in SDN-cloud using POX controller. Telecommun. Syst. 77(1), 47–62 (2021). https://
doi.org/10.1007/s11235-020-00747-w
7. Gaurav, A., et al.: Identity-based authentication mechanism for secure information sharing in
the mari-time transport system. IEEE Trans. Intell. Transp. Syst. (2021)
8. Nguyen, G.N., et al.: Secure blockchain enabled cyber–physical systems in healthcare using
deep belief network with ResNet model. J. Parallel Distrib. Comput. 153, 150–160 (2021)
9. Zhou, Z., et al.: A fine-grained access control and security approach for intelligent vehicular
transport in 6g communication system. IEEE Trans. Intell. Transp. Syst. (2021)
10. Dahiya, A., Gupta, B.B.: Multi attribute auction based incentivized solution against DDoS
attacks. Comput. Secur. 92, 101763 (2020)
11. Cviti´c, I., et al.: Boosting-based DDoS detection in internet of things systems. IEEE Internet
of Things J. 9, 2109–2123 (2021)
12. Dahiya, A., et al.: A reputation score policy and Bayesian game theory based incentivised
mechanism for DDoS attacks mitigation and cyber defense. Future Generation Computer
Systems (2020)
13. Han, J., et al.: Data Mining: Concepts and Techniques. Elsevier (2011)
14. DDoS 2019 | Datasets | Research | Canadian Institute for Cybersecurity | UNB. Accessed 28
Apr 2022
15. Alzahrani, R.J., et al.: Security analysis of DDoS attacks using machine learning algorithms
in networks traffic. Electronics 10(23), 2919 (2021)
16. He, Z., Zhang, T., Lee, R.B.: Machine learning based DDoS attack detection from source side
in cloud. In: Proceedings of the 2017 IEEE 4th International Conference on Cyber Security
and Cloud Computing (CSCloud), New York, NY, USA, 26–28 June 2017, pp. 114–120
(2017)
17. Aamir, M., et al.: DDoS attack detection with feature engineering and machine learning: the
framework and performance evaluation. Int. J. Inf. Secur. 18, 761–785 (2019)
18. Liu, Z., et al.: The prediction of DDoS attack by machine learning. In: Third International Con-
ference on Electronics and Communication; Network and Computer Technology (ECNCT
2021), vol. 12167, pp. 681–686. SPIE (2022)
19. Zewdie, T.G., Girma, A.: An evaluation framework for machine learning methods in detection
of DoS and DDoS intrusion. In: 2022 International Conference on Artificial Intelligence in
Information and Communication (ICAIIC), pp. 115–121 (2022)
20. Sahoo, S., et al.: Multiple features based approach for automatic fake news detection on social
net-works using deep learning. Appl. Soft Comput. 100, 106983 (2021)
21. Cviti´c, I., Perakovi´c, D., Periša, M., Gupta, B.: Ensemble machine learning approach for
classification of IoT devices in smart home. Int. J. Mach. Learn. Cybern. 12(11), 3179–3202
(2021). https://doi.org/10.1007/s13042-020-01241-0
22. Gupta, B.B., et al.: Machine learning and smart card based two-factor authentication scheme
for pre-serving anonymity in telecare medical information system (TMIS). Neural Computing
and Applications, 1–26 (2021). https://doi.org/10.1007/s00521-021-06152-x
23. Yamaguchi, S., Gupta, B.: Malware threat in Internet of Things and its mitigation analysis.
In: Research Anthology on Combating Denial-of-Service Attacks, pp. 371–387. IGI Global
(2021)

50 A. Mishra and D. Rawat
24. Perakovi´c, D., et al.: A Big Data and Deep Learning based Approach for DDoS Detection
in Cloud Computing Environment. In: 2021 IEEE 10th Global Conference on Consumer
Electronics (GCCE), pp. 287–290. IEEE (2021)
25. Dahiya, A., et al.: A PBNM and economic incentive-based defensive mechanism against
DDoS at-tacks. Enterp. Inf. Syst. 16(3), 406–426 (2022)
26. Dahiya, A., et al.: A reputation score policy and Bayesian game theory based incentivized
mechanism for DDoS attacks mitigation and cyber defense. Futur. Gener. Comput. Syst. 117,
193–204 (2021)
27. Chartuni, A., et al.: Multi-classifier of DDoS attacks in computer networks built on neural
networks. Appl. Sci. 11(22), 10609 (2021)
28. Zhu, X., et al.: Prediction of rockhead using a hybrid N-XGBoost machine learning
framework. J. Rock Mech. Geotech. Eng. 13(6), 1231–1245 (2021)
29. Teles, G., Rodrigues, J.J., Rabêlo, R.A., Kozlov, S.A.: Comparative study of support vector
machines and random forests machine learning algorithms on credit operation. Software:
Practice and Experience 51(12), 2492–2500 (2021)
30. Gaurav, A., et al.: A comprehensive survey on machine learning approaches for malware
detection in IoT-based enterprise information system. Enterprise Information Systems, 1–25
(2022)
31. Pedregosa, F., et al.: Scikit-learn: machine learning in python. J. Mach. Learn. Res. 12, 2825–
2830 (2011)

Role of Internet of Things and Cloud Computing
in Education System: A Review
Ajay Krishan Gairola1,2(B)and Vidit Kumar1
1Graphic Era Deemed to Be University, Dehradun, India
ajaykrishangairola@gmail.com
2Graphic Era Hill University, Dehradun, India
Abstract. The current outbreak of the coronavirus (COVID19) pandemic has
affected education across the world. To meet the current challenges posed by
COVID19, educational institutions (schools, colleges and universities) need to be
more efficient in providing quality educational services virtually. Cloud computing
and Internet of things (IOT) are such technologies that accomplishes this. In this
work, we review the recent works related to the cloud technology and IOT in
education system and explores its various benefits and challenges. Furthermore,
this article examines recent work on the potential scope of IoT in the Education
Sector.
Keywords: Cloud computing ·Virtual learning ·Internet of Things ·Online
teaching ·Virtual classes ·Smart device
1 Introduction
The international quarantine imposed since December 2019 due to the coronavirus pan-
demic has put pressure on us to reconsider the idea of e-learning [1]. Cloud computing
is a must-have within the academic process, and it is widely employed in enterprises.
For individuals in every aspect of the process of mastery, cloud technology makes train-
ing a true and enjoyable pleasure. With the help of smart devices, students can now
communicate with each other or with teachers and experience the flexibility of learning.
The offerings of cloud computing have advanced the results of an institution’s study
and have allowed professors and college students to access this modern technology as
well as receive additional blessings. Over time the educational system has changed and
at the same time is no longer limited to blackboard classrooms and textbooks. Training
in the cutting edge landscape of cloud computing has emerged as an advantage for the
commercial enterprise, while all and sundry are trying their hardest to combat the virus
posed by COVID-19 [18]. From preserving scholarly data to storing information, from
online training systems to advanced study analysis, it has completely revolutionized
teaching-learning training. Students, professors, and instructors can now enjoy cloud-
primarily based totally training‘s accessibility and convenience. The Internet of Things
connects processes, people, data, and devices, making it easier for education stakehold-
ers to convert data acquired from portable devices and sensors into useful information
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 51–60, 2022.
https://doi.org/10.1007/978-3-031-22915-2_5

52 A. K. Gairola and V. Kumar
and to take meaningful steps taken in response to that facts [2]. It is crucial to con-
sider the influence of IoT adoption in order to understand the problems and benefits
of Internet of Things in Education, especially as IoT is still in its early stages in the
education system. The Internet of Things provides numerous advantages, including: the
development of intelligent interactive classes; the ability to customize dynamic models
in which Students are active learners process; the encouragement of imagination; as well
as real-time monitoring of students’ cognitive processing. The COVID19 epidemic has
put a strain on both research and the use of new technology in education. Higher interest
in this study issue is indicated by an increase in the number of publications on the use
of IoT in education, while contemporary educational practices are a factual evidence of
such interest. Nevertheless, due to the fact that those possibilities are limitless within a
cloud pack- age and educational process, this research focuses on providing the recent
progress related to the cloud technology and IOT in education system along with its
benefits and challenges.
1.1 Cloud Computing and Education
Cloud computing is a method of delivering a variety of services via virtual machines
that are placed on top of a big pool of actual equipment in the cloud [3]. Services
are stored in laboriously scalable information in visible form centers fashionable the
cloud and accessed via the internet by some connected scheme. In the dispersed cloud
environment, we have a lot of compute power and storage capacities. Some applications
of cloud technologies in education are depicted in Table 1.
Tabl e 1 . Examples of the application of cloud technologies in education
Google Classroom [4]Google Classroom is a cloud-primarily based totally gaining
knowledge of control machine this is a part of the Google Apps for
Education suite of products. Students can use Google Classroom on
PCs, tablets, and cell-phones
Blackboard [5] Education, mobility, communication, and trade software program, in
addition to associated services, are furnished via way of means of
Blackboard to customers together with instructional institutions,
enterprises, and authorities agencies. In January 2014, round 17,000
faculties and businesses in a hundred nations had been the use of its
software program and services
Knowledge Matters [6]Knowledge Matters is a major virtualized online firm that teaches
important business principles to college and high school students
through interactive web, game-like business simulations
Coursera [7] The most well-known educational site, in my opinion. Anyone can
study a wide range of subjects on Coursera. There isn‘t a single student
in the United States, Canada, Thailand, Russia, or Ukraine who isn’t
aware of Coursera’s opportunity to gain valuable knowledge
(continued)

Role of Internet of Things and Cloud Computing in Education System 53
Tabl e 1 . (continued)
Microsoft Education Centre [8] The Microsoft Education Centre turned into supposed to permit
college students to retain studying irrespective of their circumstances.
They make on- line studying viable and offer the best training to each
and every student
Classflow [9]Classflow is an interactive screen-based course delivery program that
runs in the cloud. They offer customers unlimited access to lessons and
learning tools without a subscription
1.2 IOT Enabled Education Environment
The Internet of Things is changing the way we live by transforming every product
becoming an intelligent entity. All of this is correct in the teaching institution, where a
veritable cycle of power is intelligently carried from Smart University, Smart Classroom,
Smart Learning, Smart Learning and Smart Teaching to Smart Analysis (Table 2).
Tabl e 2 . Smart education
Smart Education The purpose of smart education is to educate students with the skills and expertise
they need to succeed in today’s market. Smart education’s success is dependent on
sensing devices, an IoT infrastructure, communication linkages, and user apps. The
IoT integration in the classroom institution would conclusion in higher educational
quality since students will learn quickly and teachers will be prepared to carry forward
their educational duties more efficiently [10]
Smart University A smart university combines cutting-edge hardware and software, cutting-edge
concepts, education techniques based on trendy, learning tactics, and smart teaching
and smart classrooms equipped with cutting-edge technology [11]. A smart university
has access to a diverse range of worldwide materials, an interactive teaching
environment that can be examined inside the network, and learning that is flexible to
data acquired. Many institutions have IoT devices such as temperature control devices,
security cameras, electricity, heating systems, and building access devices
Smart Classroom A smart classroom is a location where students can access educational activities
utilizing electronic equipment such as internet-connected gadgets, digital screens and
video projectors [12]. Beginning in 2012, a smart class is built on automated
communication devices, mobile learning and mobile technologies, which use cameras,
facial recognition algorithms, video projectors, sensors, and extra modules to keep
track of many characteristics of the natural environments. When machines are linked
to the Internet of Things, they form a smart class that allows access to knowledge from
everywhere and at any time. A smart class offers numerous advantages, including
greater information communication, flexibility, interactive learning, educational
content exchange, and improved thinking capacities
(continued)

54 A. K. Gairola and V. Kumar
Tabl e 2 . (continued)
Smart Teaching The manner in which information is transmitted via electronic devices can differ
greatly from traditional teaching approaches. The material is always accessible, and
Learning is adaptable, allowing you to stay up to date on the most recent
advancements. The Internet of Things may provide access to the actual world, which
might make teaching difficult because it must be adjusted and adapted to meet the
needs of students with various impairments. Teaching methods must also be modified
to accommodate students with disabilities
Smart Learning Smart learning is a learning approach that makes use of electronic gadgets. According
to [13], smart learning is a procedure that assists students in learning by focusing on
the subject as well as the students themselves. This technology’s intelligence,
adaptability, and efficacy are dependent on the ICT infrastructure. The usage of
Internet of things e-learning apps is critical for establishing a virtual classroom and a
competitive learning process, both worldwide and locally. Because Students have
access to every library or lab throughout the universe to collect data, conduct
experiments, and send assignment or for self-evaluation and be assigned, the Internet
of Things fosters online self-teaching
Smart Assessment Smart assessment [14] goes beyond the traditional methods by adding other types of
evaluation, such as interviews and focus groups. To make an accurate assessment, we
must consider the effects of modern technology on how we work. The evaluation
process then evolves as we engage inside an ever-expanding IoT ecosystem. To
capture student behavior in online learning assessments, new learning systems must
have the appropriate technologies. The Internet of Things instruments are available for
use in assess the student’s concentration, which is critical in assessing their education.
It is feasible to design adaptive exams that are adjusted to the student’s responses to
questions and are presented in the student’s preferred learning style. This type of
examination would allow us to delve into the students’ knowledge, how they
understand and apply it, as well as their learning styles and skills. The usage of
simulations during educational activities is an important component of smart
assessment and can also be utilized as a learning approach
2 Literature Review
2.1 Cloud Computing in Education
In recent years, cloud computing implementations have attracted attention in several
areas, including higher education in emerging markets. In this section we present a
review of the adoption of cloud computing on education. Moodle was investigated [15]
as a case of cellular cloud gaining knowledge of structures in better learning. Posted in
[16] an overview on the use of mobile cloud computing (MCC) within the instructional
field, which summarized the demanding conditions and troubles that MCC requires to
gain knowledge of structures, as well as privacy, interoperability arise because of. The
cognitive load on college students due to exceptional operating platforms, information
integrity, community availability and community speed, and large learning materials
and courses. Cloud computing (CC) in education was evaluated from the perspective
of teaching staff and IT professionals in Saudi Arabia [17]. [19] The authors studied
characteristics influencing adoption by collecting data from a mobile cloud learning
environment from Blackboard users at the University of Leeds in the United Kingdom

Role of Internet of Things and Cloud Computing in Education System 55
using a structured questionnaire. [20] The authors examined their country’s potential to
transition to remote mastering and reviewed structures that have been used in schooling
and supported by government access, as well as modern online conversation structures.
Those who are advanced with the help of using Microsoft Teams. [21] The authors
noted China’s revelations about the duration of the covid-19 lockdown with continued
learning. Authors described technical support for instructors and knowing help for col-
lege students. [22] The authors examined the risks of using an automated machine and
provided a web multi-element authentication test method. The authors [23] summarized
the conditions of seeking to achieve the amazing distance of knowing and e-checking
from the thoughts of professors and college students in Arab universities. Universities,
schools and various educational institutions are important for the standard development
of a country.
[24] Used the ambiguous AHP to take a look at the determinants of CC adoption
in better Indian educational institutions. The maximum important factors decided to be
relative profit, IT demand and security. [25] Using a SEM approach, the authors evaluated
cloud computing-based education in Saudi Arabia and influenced characteristics such
as reliability, social impact, information quality and ease of use. Tuan suggested that
it would be more accurate to assess teacher research productivity using an integrated
multi-criteria decision-making approach (MCDM). To do this, the researchers used the
hybrid AHPTOPSIS technology. Due to the suspension of on-campus classes, a large
jump in student ranks, monodemic content, and the content offered, and the material
provided, e-mastering structures have grown at an exponential rate [26,27]. The cloud
era is now being used by many educational institutions, and it is very clear [28] that it
has a shiny destiny.
In addition, due to the fact that there is a single database for all customers in the
cloud, cyber security modifications can be analyzed and made quickly [29]. [30], as it
was designed to allow customers to collaborate from anywhere at any time. It can reach
out to more students outside the general study room and meet their needs. Due to better
calls to keep education afloat, establishments are paying additional interest on a mix of
cloud generation and e-learning. Almost all educational institutions saw it as a viable and
suitable e-learning option. Nevertheless, the lack of study may also provide a theoretical
framework on which to build a technology. On the other hand, the potential inherent in
the cloud approach can also be highlighted as a major advantage in the development of
an analytical framework and one-hit training techniques [31]. However, in the literature,
common features of the cloud are associated with social participation and collaborative
learning activities [32].
2.2 IOT in Education
The term “Internet of Things” (IoT) refers to state-of-the-art technology that connects all
intelligent objects in a network without the need for human intervention. This is indeed
a new study focus that would have recently discovered an important and compelling
research base in a wide range of academic and industrial disciplines [33]. According to
Walcott [34], many governments are implementing the latest digital defense strategies in
the fight against COVID 19. During this time, digital technology and innovation gradually
became the focus of mankind. The economic demands of COVID19 strongly drive the

56 A. K. Gairola and V. Kumar
deployment and creation of new digital technologies at a particular pace and scale.
Population surveillance, response assessment, incident detection and touch tracking is
one of the digital tools used to facilitate the international public health impact of COVID
19, with a focus on public participation and data mobility. According to Islam [35], the
integration of IoT with advanced technology could be a major step forward in efforts to
combat new epidemics. The potential of the Internet of Things will have a significant
impact on the ability of Western countries to achieve the SDGs (Sustainable Development
Goals). In environments where Internet-ofThings-enabled devices and applications are
used, it is essential to implement specified protocols, patient monitoring and primary
identification procedures, to reduce the chances of spreading the coronavirus.
According to Jawed [36], the Internet of Things can send and receive both information
and physical goods (IoT). Intelligent hospital equipment and concepts were controlled
via wireless and wired Internet. Various medical diagnostics, instruments, advanced
imaging equipment, artificial intelligence and sensors are essential for the implementa-
tion of IoT in the medical field. Intelligent technologies can collect and share data to
carry out essential tasks in our daily lives. The application will pave the way for enter-
tainment systems, automobiles, connected healthcare and smart cities. These advances
have increased both the quality of life and the efficiency of industries and societies, both
new and old. This technology is flourishing in health surveillance during the COVID 19
pandemic. According to Nasajpur et al. [37] Innovation has retained most of the infor-
mation about COVID19 patients inside the data center to ensure adequate attention,
and that could be more helpful. Internet of Things (IoT) combines all-digital, computing
technologies and mechanicals to transmit data over the web without human intervention.
In this dire scenario, many people die of late and incorrect medical information. The
Internet of Things is taking over every day human activities and changing health prob-
lems. Sensors are used to quickly notify the system of health issues [38]. The successful
operation of medical institutions requires proper equipment. During the COVID-19 pan-
demic, the use of the Internet of Things improves patient care. Smart medical devices
are connected via smart connectors to deliver important medical data to doctors. These
devices use the Internet of Things to successfully track real-time data, saving lives from
a variety of health problems. The Internet of Things (IoT) has great potential to analyze
and leverage impactful activities including after-services [36]. De Rauer and Radanlive
[39,40] focused on ethical IoT design updates and IoT design, but did not discuss the
implications of the coupled and multiple risks of IoT system-themes. They concluded
that before new ICT systems are incorporated, production facilities should be coupled
with an ethical assessment of cyber threats. They enable governments, health experts, and
medical organizations to build a framework to provide guidance in this article [41]asthe
introduction of IoT into the vaccination supply chain increases risk. [42] Applications of
the Internet of Things include contact tracing devices, wearable health monitors, thermal
cameras, temperature sensors and package tracking to help fight disease by providing
critical data needed for the safe delivery of COVID19 vaccines. In this COVID situa-
tion, IoT has helped to make automated activities in warehouses and supply chains more
resilient to encourage social distancing and secure remote access to industrial machines.
By studying the potential of IoT in the socio-economic development areas of
Bangladesh, Parvez et al. [43] Created a conceptual framework model, and the model

Role of Internet of Things and Cloud Computing in Education System 57
showed that Bangladesh needed to develop a set of policies for IoT deployment to
implement a national strategy on the Internet of Things. Miyazi et al. [44]TheIoT,
introduced in Bangladesh, reveals technical challenges, financial challenges, security,
privacy issues and device reliability, along with opportunities such as occupational safety,
mHealth, traffic safety, service management and environmental monitoring. Sarkar et al.
[45] Highlighting the future prospects and problems of some of the most promising IoT
applications. As per the literature review on IoT applications in Bangladesh, no such
in-depth work has been found on the current scenario of employing IoT in various indus-
tries during COVID-19. As a result, a conceptual model of the impact of Inter-Net of
Things applications across multiple industries was created during the pandemic. Dur-
ing the pandemic, this study looks at the barriers and benefits of adopting IoT services
across multiple sectors, and the findings will help organizations respond and adapt to
IoT services more quickly, giving them a competitive advantage.
3 Discussion and Conclusion
In India, cloud computing adaptation in higher education is an under-studied area and
the literature does not document systematic studies. We studied in this article how cloud
computing can be used in educational contexts. Due to the COVID-19 pandemic which
has prompted many schools, colleges and establishments to supply online training, it
has become mandatory. According to the analysis’ overview, using cloud services in
E-learning is a good option since it allows teachers to take use of cloud adaptability,
flexibility, and security to reflect the primary framework of E-learning education acces-
sible from anywhere, at any time, and on any device. We can fully use the prospects
presented by an efficient learning environment with specialized information that is easily
adaptable to today’s educational paradigm. Integrating an elearning system into the cloud
has several advantages, including increased storage, computing, network connectivity
and prioritize software and hardware cost savings. On the other hand, it offers a more
diverse range of educational programs at a lower license cost. The replacement rate for
student computers is lower due to the extended machine life. These savings add up to a
reduction in IT personnel costs associated with computer lab maintenance and software
updates. Today’s e-learning services and systems are all about personalizing learning and
learning for each user. As a result of this technology, students receive generic e-learning
that is not tailored to their specific needs. In most modern systems, interaction between
professors and students is essential for improving the quality of each student’s learning
experience. When evaluating the scale of a problem, there are many things to consider.
In response to customer concerns about security and privacy, cloud service providers
have made major investments in cloud infrastructure and platforms. Furthermore, coun-
try limits are necessary since some countries require data to be maintained within their
borders, making data storage remotely or outside of the country illegal. As per current
research, academics have a wealth of data at their disposal to aid in building cloud-based
elearning frameworks and implementations. A quantitative assessment of the impact of
switching to a cloud e-learning environment on several factors such as access speed,
educational quality, and return will be conducted in the future. The adoption of IoT in
universities may be influenced by education policy in terms of administrative support

58 A. K. Gairola and V. Kumar
and change mindset. There is a need to examine the advantages and disadvantages of
Internet of Things in depth. Information and communication technology (ICT), a soci-
ety that places a high value on acquiring knowledge, and the current pandemic have all
contributed to an increase in the amount of pressure on the education system to adopt
ICT and make education more intelligent. There is also a need to explore machine learn-
ing algorithms [46–50] in cloud based analysis of education systems for tasks such as
student monitoring, student lecture engagement, etc.
References
1. Dias, S.B., Hadjileontiadou, S.J., Diniz, J., Hadjileontiadis, L.J.: DeepLMS: a deep learning
predictive model for supporting online learning in the Covid-19 era. Sci. Rep. 10(1), 1–17
(2020)
2. Bagheri, M., Movahed, S.H.: The Effect of the Internet of Things (IoT) on Education Business
Model, in Proc, pp. 435–441. SITIS, Naples, Italy (2016)
3. Gong, C., Liu, J., Zhang, Q., Chen, H., Gong, Z.: The characteristics of cloud computing. In:
2010 39th International Conference on Parallel Processing Workshops, pp. 275–279. IEEE
(2010)
4. https://classroom.google.com/. Accessed 26 May 2022
5. https://www.blackboard.com/en-apac. Accessed 26 May 2022
6. https://knowledgematters.com/. Accessed 26 May 2022
7. https://www.coursera.org/. Accessed 26 May 2022
8. https://education.microsoft.com/en-us. Accessed 26 May 2022
9. https://classflow.com/. Accessed 26 May 2022
10. Mohanty, D.: Smart learning using IoT. Int. Res. J. Eng. Tech. 6(6), 1032– 1037 (2019)
11. Uskov, V.L., Bakken, J.P., Howlett, R.J., Jain, L.C. (eds.): SEEL 2017. SIST, vol. 70. Springer,
Cham (2018). https://doi.org/10.1007/978-3-319-59454-5
12. Pai, S.S., et al.: IOT application in education. Int. J. Adv. Res. Ideas Innovations Technol.
2(6), 20–24 (2017)
13. Gwak, D.: The meaning and predict of smart learning. In: Proceedings of the Smart Learning
Korea (2010)
14. Aljohany, D.A., Mohamed, R., Saleh, M.: ASSA: adaptive E-learning smart students
assessment model. Int. J. Adv. Comput. Sci. Appl. 9(7), 128–136 (2018)
15. Wang, M., Chen, Y., Khan, M.J.: Mobile cloud learning for higher education: a case study of
moodle in the cloud. Int. Rev. Res. Open Distrib. Learn. 15(2), 254–267 (2014)
16. Sarode, N., Bakal, J.W.: A review on use of mobile cloud system in educational sector. In:
2020 6th International Conference on Advanced Computing and Communication Systems
(ICACCS), pp. 715–720. IEEE (2020)
17. Almutairi, M.M.: A review of cloud computing in education in Saudi Arabia. Int. J. Inform.
Technol. 12(4), 1385–1391 (2020). https://doi.org/10.1007/s41870-020-00452-6
18. Kumar, V.: A review on deep learning based diagnosis of COVID-19 from X-ray and CT
images. In: 2022 International Mobile and Embedded Technology Conference (MECON),
pp. 547–552. IEEE (2022)
19. Sultana, J.: Determining the factors that affect the uses of mobile cloud learning (MCL)
platform blackboard-a modification of the UTAUT model. Educ. Inform. Technol. 25(1),
223–238 (2020). https://doi.org/10.1007/s10639-019-09969-1
20. Basilaia, G., Kvavadze, D.: Transition to online education in schools during a SARS-CoV-2
coronavirus (COVID-19) pandemic in Georgia. Pedagogical Research 5, 4 (2020)

Role of Internet of Things and Cloud Computing in Education System 59
21. Huang, R.H., Liu, D.J., Tlili, A., Yang, J.F., Wang, H.H.: Handbook on Facilitating Flexible
Learning During Educational Disruption: The Chinese Experience in Maintaining Undis-
rupted Learning in COVID-19 Outbreak, pp. 1–54. Smart Learning Institute of Beijing Normal
University, Beijing (2020)
22. Mallik, S., Halder,S., Saha, P., Mukherjee, S.: Multi-factor authentication-based E-exam man-
agement system (EEMS). In: Bhattacharjee, D., Kole, D.K., Dey, N., Basu, S., Plewczynski,
D. (eds.) Proceedings of International Conference on Frontiers in Computing and Systems.
AISC, vol. 1255, pp. 711–720. Springer, Singapore (2021). https://doi.org/10.1007/978-981-
15-7834-2_66
23. Bashitialshaaer, R., Alhendawi, M., Lassoued, Z.: Obstacle comparisons to achieving distance
learning and applying electronic exams during COVID19 pandemic. Symmetry 13(1), 99
(2021)
24. Sharma, M., Gupta, R., Acharya, P.: Factors influencing cloud computing adoption forhigher
educational institutes in India: a fuzzy AHP approach. Int. J. Inf. Technol. Manage. 19(2–3),
126–150 (2020)
25. Naveed, Q.N., Alam, M.M., Qahmash, A.I., Quadri, K.M.: Exploring the determinants of
service quality of cloud E-learning system for active system usage. Appl. Sci. 11(9), 4176
(2021)
26. Khan, R.M.I., Radzuan, N., Farooqi, S., Shahbaz, M., Khan, M.: Learners’ perceptions on
whatsapp integration as a learning tool to develop EFL spoken vocabulary. Int. J. Lang. Educ.
5(2), 1–14 (2021)
27. Khan, R.M.I., Shahbaz, M., Kumar, T., Khan, I.: Investigating reading challenges faced by
EFL learners at elementary level. Register J. 13(2), 277–292 (2020)
28. Khan, I., Ibrahim, A.H., Kassim, A., Khan, R.M.I.: Exploring the EFI learners’ attitudes
towards the integration of active reading software in learning reading comprehension at tertiary
level. MIER J. Educ. Stud. Trends Pract., 248-266 (2020)
29. Bhardwaj, A., Goundar, S.: A framework to define the relationship between cyber security
and cloud performance. Comput. Fraud & Secur. 2019(2), 12–19 (2019)
30. Kaisara, G., Bwalya, K.J.: Investigating the e-learning challenges faced by students during
COVID-19 in Namibia. Int. J. High. Educ. 10(1), 308–318 (2021)
31. Park, J.H., Park, J.H.: Blockchain security in cloud computing: use cases, challenges, and
solutions. Symmetry 9(8), 164 (2017)
32. Marinescu, D.C.: Cloud Computing: Theory and Practice. Morgan Kaufmann (2017)
33. Mohammed, T., Jean-Yves, C., Peter, B., Christophe, R.: Petrogenesis of the post-collisional
Bled M’Dena volcanic ring complex in Reguibat Rise (western Eglab shield, Algeria). J. Afr.
Earth Sci. 166, 102250 (2020)
34. Walcott, D.A.: How the fourth industrial revolution can help us beat COVID-19. In: World
Economic Forum (2020). https://www.weforum.org/agenda/2020/05/how-the-fourth-indust
rialrevolution-can-help-us-handle-the-threat-of-covid-19
35. Islam, A., Anum, K., Dwidienawati, D., Wahab, S., Abdul, L.A.: Building a post COVID-19
configuration between Internet of Things (IoT) and sustainable development goals (SDGs)
for developing countries. J. Arts Soc. Sci. 4(1), 45–58 (2020)
36. Javaid, M., Khan, I.H.: Internet of Things (IoT) enabled healthcare helps to take the challenges
of COVID-19 pandemic. J. Oral Biol. Craniofac. Res. 11(2), 209–214 (2021)
37. Nasajpour, M., Pouriyeh, S., Parizi, R.M., Dorodchi, M., Valero, M., Arabnia, H.R.: Internet
of Things for current COVID-19 and future pandemics: an exploratory study. J Healthcare
Inf Res. 1, 40 (2020)
38. Fahrni, S., Jansen, C., John, M., Kasah, T., Körber, B., Mohr, N.: Coronavirus: Industrial IoT
in Challenging Times. McKinsey & Company, New York (2020)

60 A. K. Gairola and V. Kumar
39. Radanliev, P., De Roure, D.: Alternative mental health therapies in prolonged lockdowns:
narratives from Covid-19. Heal. Technol. 11(5), 1101–1107 (2021). https://doi.org/10.1007/
s12553-021-00581-3
40. Radanliev, P., De Roure, D.: Epistemological and bibliometric analysis of ethics and shared
responsibility—health policy and IoT systems. Sustainability. 13(15), 8355 (2021)
41. Radanliev, P., De Roure, D., Ani, U., Carvalho, G.: The ethics of shared Covid-19 risks: an
epistemological framework for ethical health technology assessment of risk in vaccine supply
chain infrastructures. Heal. Technol. 11(5), 1083–1091 (2021). https://doi.org/10.1007/s12
553-021-00565-3
42. Forum, W.E.: State of the Connected World (2020). http://www3.weforum.org/docs/WEF_
The_State_of_the_Connected_World_2020.pdf
43. Parvez, N., Chowdhury, T.H., Urmi, S.S., Taher, K.A.: Prospects of Internet of Things
for Bangladesh. In: 2021 International Conference on Information and Communication
Technology for Sustainable Development (ICICT4SD), pp. 481–485 (2021)
44. Miazi, M.N.S., Erasmus, Z., Razzaque, M.A., Zennaro, M., Bagula, A.: Enabling the Internet
of Things in developing countries: opportunities and challenges. In: 2016 5th International
Conference on Informatics, Electronics and Vision (ICIEV), pp. 564–569. IEEE (2016)
45. Sarker, S., Roy, K., Afroz, F., Pathan, A.-S.: On the opportunities, applications, and chal-
lenges of internet of things. In: Khan, M.A., Quasim, M.T., Algarni, F., Alharthi, A. (eds.)
Decentralised Internet of Things. SBD, vol. 71, pp. 231–254. Springer, Cham (2020). https://
doi.org/10.1007/978-3-030-38677-1_11
46. Kumar, V., et al.: Hybrid spatiotemporal contrastive representation learning for content-based
surgical video retrieval. Electronics 11, 1353 (2022)
47. Kumar, V., Tripathi, V., Pant, B.: Learning unsupervised visual representations using 3d
convolutional autoencoder with temporal contrastive modeling for video retrieval. Int. J.
Math. Eng. Manag. Sci. 7(2), 272–287 (2022)
48. Kumar, V., Tripathi, V., Pant, B.: Enhancing unsupervised video representation learning by
temporal contrastive modelling using 2d CNN. In: 5th IAPR International Conference on
Computer Vision & Image Processing (CVIP 2021)
49. Kumar, V., Tripathi, V., Pant, B.: Unsupervised learning of visual representations via rotation
and future frame prediction for video retrieval. In: Singh, M., Tyagi, V., Gupta, P.K., Flusser,
J., Ören, T., Sonawane, V.R. (eds.) ICACDS 2021. CCIS, vol. 1440, pp. 701–710. Springer,
Cham (2021). https://doi.org/10.1007/978-3-030-81462-5_61
50. Kumar, V., Tripathi, V., Pant, B.: Exploring the strengths of neural codes for video retrieval. In:
Tomar, A., Malik, H., Kumar, P., Iqbal, A. (eds.) Machine Learning, Advances in Computing,
Renewable Energy and Communication. LNEE, vol. 768, pp. 519–531. Springer, Singapore
(2022). https://doi.org/10.1007/978-981-16-2354-7_46

Smart Communication and Technology

An Effective Image Augmentation Approach
for Maize Crop Disease Recognition
and Classification
M. Nagaraju1, Priyanka Chawla2(B), and Rajeev Tiwari3
1School of Computer Science and Engineering, Lovely Professional University, Phagwara,
Punjab, India
2Department of Computer Science & Engineering, National Institute of Technology Warangal,
Hanamkonda, Telangana, India
priyankac@nitw.ac.in
3Systemic, School of Computer Science, University of Petroleum and Energy Studies,
Dehradun, Uttarakhand, India
Abstract. Deep learning techniques have been applied to computer vision appli-
cations like image recognition and classification successfully. Especially, con-
volutional neural networks preserve the characteristics of an object in an image
using kernels and performs recognition very efficiently. However, the performance
of these networks depends on larger datasets which is a big challenge to the
researchers in the agriculture field. Image augmentation can be a better solution
that supports the neural network model to perform the classification task efficiently
with more input images. In this paper, severalimage augmentation techniques were
applied to generate varieties of new images from the original image. The paper
proposes a new CNN-based model for the classification of six diseased and one
healthy maize crop images. The proposed model will be trained for twice indepen-
dently with 4652 original dataset images and 10640 augmented images dataset.
Finally, the outcomes will be analyzed separately with respect to precise and loss
functions. Before the implementation of augmentation approach, the proposed
model has achieved 99.61% training and 77.44% classification accuracies and
does not control overfitting. Moreover, after applying augmentation techniques,
the model has obtained 95.96% of training accuracy and 93.61% of classification
accuracy and controlled overfitting. Therefore, it has been proved that the image
augmentation approach and the proposed convolutional neural network model
contribute a better solution while classifying maize crop diseases with a higher
level of accuracy.
Keywords: Computer vision ·Convolutional neural networks ·Deep learning ·
Image augmentations ·Image classification ·Image preprocessing techniques
1 Introduction
A practical approach to deal with less images in computer vision is typical. Augment
the training images artificially can extract new images and may reduce overfitting [8].
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 63–72, 2022.
https://doi.org/10.1007/978-3-031-22915-2_6

64 M. Nagaraju et al.
Imbalanced datasets can also be solved by applying augmentation techniques to transfer
the original shape of the plant generating additional images [9]. Image augmentations
gives a resultant dataset that is six times greater than the quantity of original set. The
proposed model LeafGAN boosted the accuracy by 7.4% [2]. Model generalization can
be improved by performing image preprocessing and augmentation. The experiments
were conducted to evaluate the efficiency of the proposed model to classify DiaMOS
plant dataset [3].
An advanced machine learning (ML) model is proposed to classify the major diseases
in banana crop using ariel images. The results obtained by the trained models proposed
that image augmentation have given positive outcomes on disease classification. The
model has a control on training rate without any overfitting [4]. An image augmentation
strategy has been followed to amplify the original images and tested the performance
classification of apple. The results have shown that the proposed model achieved 6.3%
higher recognition accuracy [3]. An enhanced classification model is proposed to increase
the classification accuracy and address the overfitting problem. The model has achieved
24.4% higher overall classification accuracy using conventional image augmentations
[1]. Synthetic image is the most usual method of data augmentation. The proposed
model achieved 7% higher improved accuracy over the existing models [5]. A transfer
learning concept has implemented by modifying VGG-16 to classify the images obtained
from different mango farms. Image augmentation process is adopted and achieved 73%
accuracy on the training dataset and 73% on the testing set. Data augmentation leads
to 13.43% improvement on the testing data [7]. A conditional deep neural network is
proposed for vigor rating of plants and investigated that after data augmentations the
model has improved the classification accuracy and succeeded to obtain 23% increase
in F1score. The proposed approach has resolved the problem of insufficient data size in
plant diseases task [11–14,12].
In this article, we have identified the need to expand all the seven datasets with data
augmentation techniques. Image augmentation generates the new images that allows
to balance all the disease classes with equal number of images. Firstly, an image of a
diseased leaf with RGB representation is augmented into different variations for each
transformation type. Seven types of image transformations like random rotation, hor-
izontal shift, vertical shift, horizontal flip, vertical flip, random zooming, and random
brightness are applied to an image of a diseased leaf. The outcome of these techniques
generates a set of newly generated set of images that are used further to train a convolu-
tional neural network (CNN). The novelty of the proposed work is to generate the new
images dataset, design a CNN model, perform disease classification with original and
augmented datasets.
The other sections are organized as: Sect. 2discusses the approach followed to
apply the image augmentation techniques. Section 3describes the experimental results
achieved from the image augmentation approach followed by the conclusion in Sect. 4.
2 Image Augmentation Approach
Image Augmentation or simple IA is an essential approach that replicates the given
image with few transformations. The augmentation technique increases the diversity of

An Effective Image Augmentation Approach for Maize Crop Disease Recognition 65
images by changing each image in different ways like rotating, shifting, zooming, and
flipping.
In this paper, a supervised learning named RGB Image Augmentation Approach
(IAA) is followed to generate new images. The approach directs to improve the tradi-
tional techniques like augmentation by considering the requirement for new augmenta-
tion techniques. Figure 1shows some original images and the images generated after
applying the augmentations. The study employs the IAA as a preprocessing procedure
to identify the converted images automatically. The newly generated images serve as
the preprocessing procedure to feed the convolutional neural networks with the input
images for training.
Fig. 1. Sample Images (a) original (b) augmented.
3 Experimental Results and Observations
A CNN model is developed to implement the IA approach and perform the image
classification in this section.
3.1 RGB Image Augmentation
A new CNN model is developed with four convolution and max-pooling layers one and
all continued with 1 flatten and 2 dense layers. Figure 2shows the summary of the
proposed model and the related hyperparameters are shown in Fig. 3.
3.2 CNN Evaluation
The classification performance of the proposed CNN model is evaluating by conducting
the experiments with 4652 leaf images collected from Kaggle repository (Kaggle Dataset
n.d.). These images belong to seven different classes of maize crop. The image dataset is
split into two subsets as training set and testing set. IAA will be applied on the training
images to generate new images with different variants. After applying IAA, the quantity
of the training dataset is increased to 10640 images. The testing dataset is prepared with

66 M. Nagaraju et al.
Fig. 2. Proposed model summary
Fig. 3. The proposed 11-layer CNN model
2660 images that are randomly selected from the first dataset of 4652 images. The list
of diseases and the number of images considered for each disease is depicted in Table
1. The model is evaluated with original and augmented datasets by training and testing
individually. Later, the results obtained are used to validate and compare the actual
predictions with respect to accuracy, loss and confusion matrix.
3.2.1 CNN Model Implementation
The present sequential model is designed with four convolution layers (Conv2D_1, 2,
3 and 4 layers) with kernel size 3 ×3 followed by four max-pooling layers with pool
size 2 and 2 strides. The model has one fully connected layer, and two dense layers with
128 units for first and 7 units for the last layers. First, the proposed model is feed with
original dataset (without augmentation) and performed the classification. Figure 4shows
the performance with training, testing accuracy and loss curves before implementing
IAA. The accuracy and loss curves are plotted individually for better understanding and
comparison of results.

An Effective Image Augmentation Approach for Maize Crop Disease Recognition 67
Tabl e 1 . Details of training and testing datasets
Disease class 1 2 3 4 Total images
Anthracnose Leaf Blight - ALB 435 380 1520 380 2280
Anthracnose Stalk Rot – ASR 512 380 1520 380 2280
Eye Spot – ES 352 380 1520 380 2280
Gabriella Stalk Rot – GSR 452 380 1520 380 2280
Health – H 1320 380 1520 380 2280
Northern Corn Leaf Spot –NCLS 1170 380 1520 380 2280
Southern Rust - SR 411 380 1520 380 2280
Tota l 4652 2660 10640 2660 13300
1-Number of training images before IAA, 2-Number of testing image before IAA, 3-Number of
training images after IAA, 4-Number of testing images after IAA.
Fig. 4. Performance evaluation before applying the IAA technique
The plots in Fig. 4(a) shows the variation in learning of the proposed model using
training and testing accuracies. The plots in Fig. 4(a) shows the variation in learning of
the proposed model using training and testing losses. ‘Acc’ and ‘val_acc’ indicates the
training and testing accuracy curves whereas ‘loss’ and ‘val_loss’ indicate the training
and testing loss curves. The large gap between the accuracies and losses curves, a higher
difference of 22.17% among the training and testing accuracies describe that the model
is overfit to the training dataset and not performed the classification well on the testing
set. Before applying IAA and after running the model using 100 epochs, the training
accuracy of 99.61%, and the testing accuracy of 77.44% are obtained.
In addition to confusion metrics, the other performance metrics like precision, recall,
and F1-score have also computed and the values are presented in Table 2. The confusion
matrix of the proposed model shown in Fig. 5(a) and 5(b) revealed that the classification
exhibited more than 95% accuracy in only two classes ES (95.7%) and H (97.8%).
The precision metric values for ES and H disease classes are found as 93% and 91%
respectively. The recall metric values for ES and H disease classes are found as 96%
and 98% respectively. The F1-score metric value is reported as 96% for H disease class.

68 M. Nagaraju et al.
Tabl e 2 . Classification performance before and after applying IAA
Class # 1 2 1 2 1 2 Support
Precision Recall F1-score
00.61 0.89 0.71 0.95 0.65 0.92 380
10.87 0.97 0.89 0.90 0.88 0.93 380
20.78 0.94 0.96 0.94 0.86 0.94 380
30.93 0.88 0.98 1.00 0.96 0.93 380
40.91 0.95 0.71 0.94 0.80 0.95 380
50.59 0.95 0.58 0.87 0.59 0.91 380
60.77 0.98 0.59 0.95 0.67 0.97 380
Accuracy 0.77 0.94 2660
Macro Avg 0.78 0.94 0.77 0.94 0.77 0.94 2660
Weighted Avg 0.78 0.94 0.59 0.94 0.67 0.94 2660
1-Before IAA and 1-After IAA.
The metric values revealed that the proposed model has performed the classification
more efficiently only for two disease classes ES and H. It is observed that the model is
not so efficient during the classification of other five diseases ALB, ASR, GSR, NCLS
and SR. Second, the proposed model is feed with augmented dataset and performed the
classification. Figure 6shows the performance with training and testing accuracy and
loss curves after applying IAA. The accuracy and loss curves are plotted separately or
better understanding and comparison of results.
The plots in Fig. 6(a) shows the variation in learning of the proposed model using
training and testing accuracies. The plots in Fig. 6(a) shows the variation in learning of
the proposed model using training and testing losses. ‘Acc’ and ‘val_acc’ indicates the
training and testing accuracy curves whereas ‘loss’ and ‘val_loss’ indicate the training
and testing loss curves. The proposed model after training and testing with the images
generated after applying IAA has obtained 95.96% training and 93.61% testing accu-
racies. A 2.35% difference among the accuracies between training and testing datasets
shows that the model has performed well when compared the results obtained before
applying IAA. The accuracy and loss curves describe that the present model is not overfit
to the training images dataset and performed the classification well even on the testing
dataset. The results shown in Fig. 6explains that the proposed model is so efficient when
it is trained with a greater number of input images. In this regard, image augmentation
techniques have contributed a lot to increase the number of training images and achieve
better classification performance.
In addition to confusion matrix, the performance metrics like precision, recall, and
F1-score have also computed and the values are illustrated in Table 2. The confusion
matrix of the proposed model depicted in Fig. 7(a) and 7(b) revealed that the classification
exhibited 100% accuracy for H. The results exhibited more than 95% accuracy in only
two classes ALB (95.2%) and SR (95%). It is observed that the classification exhibited

An Effective Image Augmentation Approach for Maize Crop Disease Recognition 69
Fig. 5. Confusion Matrix before Implementing IAA Technique (a) without normalization (b) with
normalization (c) color scale reflections
equal or more than 90% in two disease classes ES (93.6%), and GSR (94.2%). The
classification exhibited equal or more than 95% accuracy in three classes ALB (95.2%),
H (100%), and SR (95%). Precision metric values of ASR, ES, GSR, NCLS, and SR
disease classes are found as 97%, 94%, 95%, 95% and 98% respectively. Recall metric
values of ALB, ASR, ES, H, GSR, and SR disease classes are found as 95%, 90%,
94%, 100%, 94% and 95% respectively. The F1-score is reported 100% for H disease
class. The results revealed that the proposed model has performed the classification of
six disease classes ALB, ASR, ES, H, GSR, and SR more efficiently. It is observed that
the model is not so efficient for classification of only one disease class NCLS. Figure 8
depicts the classification performance of the proposed model before and after applying
IAA.

70 M. Nagaraju et al.
Fig. 6. Performance evaluation after applying the IAA technique
Fig. 7. Confusion Matrix after implementing IAA technique (a) without normalization (b) with
normalization (c) color scale reflections

An Effective Image Augmentation Approach for Maize Crop Disease Recognition 71
Fig. 8. Performance measure before/after IAA technique
4 Conclusion
An adoptive supervised learning approach IAA to perform image augmentations to
generate the multiple images automatically in this paper. The enhanced dataset can
be adaptive to train any CNN model. The present paper has proposed a new CNN-
based model with 11 layers to perform the image classification of seven maize crop
disease classes. Later, the model has trained and tested with different datasets. First,
the model has trained with only 4652 images collected from Kaggle repository. Next,
the same model has trained with 10640 augmented images. Finally, the accuracy and
loss values have collected separately for two experiments and compared to identify
the difference in classification performances. The first experimental results show that
before applying augmentations, the proposed CNN model has obtained 99.61% training
accuracy and testing accuracy of only 77.44%. The second experimental results show
that after applying augmentation approach has obtained a training accuracy of 95.96%
and testing accuracy of 93.61%. The observations have described that the proposed CNN
model is so efficient while classifying the maize crop diseased and healthy images. CNN
model is extremely effective when it comes to classifying the diseased images of maize
crop. The findings drew attention to more advanced transformations for increasing the
number of images in training set and assisting in the prevention of model overfitting.
References
1. Bi, L., Hu, G.: Improving image-based plant disease classification with generative adversarial
network under limited training set. Front. Plant Sci. 11 (2020). https://doi.org/10.3389/fpls.
2020.583438
2. Cap, Q.H., Uga, H., Kagiwada, S., Iyatomi, H.: LeafGAN: an effective data augmentation
method for practical plant disease diagnosis (2020). http://arxiv.org/abs/2002.10100
3. Fenu, G., Malloci, F.M.: Using multioutput learning to diagnose plant disease and stress
severity. Complexity (2021). https://doi.org/10.1155/2021/6663442

72 M. Nagaraju et al.
4. Gomez Selvaraj, M., et al.: Detection of banana plants and their major diseases through aerial
images and machine learning methods: a case study in DR Congo and Republic of Benin.
ISPRS J. Photogramm. Remote Sens. 169, 110–124 (2020). https://doi.org/10.1016/j.isprsj
prs.2020.08.025
5. Hu, G., Peng, X., Yang, Y., Hospedales, T., Verbeek, J.: Frankenstein: learning deep face
representations using small data (2016). http://arxiv.org/abs/1603.06470
6. Kaggle Dataset: Corn or Maize Dataset Corn or Maize Leaf Disease Dataset | Kaggle (n.d.)
7. Kusrini, K., et al.: Data augmentation for automated pest classification in Mango farms.
Comput. Electron. Agric. 179 (2020). https://doi.org/10.1016/j.compag.2020.105842 279
8. Shorten, C., Khoshgoftaar, T.M.: A survey on image data augmentation for deep learning. J.
Big Data 6(1), 1–48 (2019). https://doi.org/10.1186/s40537-019-0197-0
9. Toda, Y., Okura, F.: How convolutional neural networks diagnose plant disease. Plant
Phenomics (2019). https://doi.org/10.34133/2019/9237136
10. Yan, Q., et al.: Apple leaf diseases recognition based on an improved convolutional neural
network. Sensors 20, 3535 (2020). https://doi.org/10.3390/s20123535
11. Zhu, F., He, M., Zheng, Z.: Data augmentation using improved cDCGAN for plant vigor
rating. Comput. Electron. Agric. 175 (2020). https://doi.org/10.1016/j.compag.2020.105603
12. Mishra, A.M., Harnal, S., Gautam, V., Tiwari, R., Upadhyay, S.: Weed density estimation in
soya bean crop using deep convolutional neural networks in smart agriculture. J. Plant Dis.
Prot., 1–12 (2022)
13. Kaur, P., et al.: Recognition of leaf disease using hybrid convolutional neural network by
applying feature reduction. Sensors 22(2), 575 (2022)
14. Kaur, P., et al.: A hybrid convolutional neural network model for diagnosis of COVID-19
using chest X-ray images. Int. J. Environ. Res. Public Health 18(22), 12191 (2021)

Implementation of Artificial Intelligence (AI)
in Smart Manufacturing: A Status Review
Akash Sur Choudhury1, Tamesh Halder2, Arindam Basak1(B),
and Debashish Chakravarty2
1School of Electronics Engineering, KIIT, Bhubaneswar, Odisha, India
arindambasak2007@gmail.com
2Department of Mining Engineering, Indian Institute of Technology, Kharagpur, West Bengal,
India
Abstract. In today’s world, artificial intelligence (AI) is widely considered one of
the highly innovativetechnologies. Usage of AI has been implemented nearly in all
sectors such as manufacturing, R&D, education, smart cities, agriculture, etc. The
new era of the Internet plus AI has resulted in the high-speed evolutionof the central
technologies, analyzed based on research regarding recent artificial intelligence
(AI) applications in smart manufacturing. It is necessary to set up an industry
that must be flexible with turbulent changes and adequately manage highly skilled
employees and workers to design a more suitable working atmosphere for both men
and technology. Google Scholar is widely used to explore several keywords and
their combinations and search and examine the relevant articles, papers, journals,
and study data for conducting this manuscript. The recent progress in intelligent
manufacturing is discussed by observing the outlook of intelligent manufacturing
technology and its applications. Lastly, the study talks about the scope of AI and
how it is implemented in today’s smart manufacturing sector of India, focusing
on its present status, limitations, and suggestions for overcoming problems.
Keywords: Artificial intelligence ·Smart manufacturing ·Industry 4.0 ·IIOT ·
CPS ·Machine learning ·Deep learning ·RUL ·ICT
1 Introduction
AI refers to technology having perceptive and psychological abilities. It has also autho-
rized first-class coherent processes such as thinking, learning, perceiving, decision-
making, problem-solving, data collection, segregation, and analysis to supplement
human brainpower. In 1956, computer scientists Allen Newell, Marvin Minsky, John
McCarthy, Arthur Samuel, and Herbert Simon developed artificial intelligence theory.
Late in the 1990s and early in the twenty-first century, AI usage is rapidly transforming
the globe, increasing the significance of analytics and enormous growth of computing
ability [1]. Figure 1tells us about the applications of AI/ML algorithms in different
processes such as fault prediction, security, etc. In Fig. 2, various machine learning clas-
sifications and their characteristics are discussed. According to the training system and
the input data type, there are three types of machine learning algorithm classifications:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 73–85, 2022.
https://doi.org/10.1007/978-3-031-22915-2_7

74 A. S. Choudhury et al.
supervised learning, unsupervised learning, and reinforcement learning [2,5].
Fig. 1. Applications of AI/ML algorithms [5].
Fig. 2. Machine learning classifications and their significant characteristics [5].
Semi-supervised learning: this method has a small set of labeled information, and
the remaining data is unlabelled hence the name semi-supervised.
Now smart manufacturing refers to the manufacturing technology aiming to provide
the industrial setting for intelligent, real-time, autonomous, and interoperable production
environments.
It integrates recent and innovative information and communication technologies
(including 5G networks and Wifi), such as the Internet of Things (IoT), cloud com-
puting (CC), and cyber-physical systems (CPS) powered by AI/ML decision-making
technologies, and results in accurate fault detection and also real-time defective product
recognition [3]. This paper describes the cycle of Industry 4.0, such as data acquisition,
monitoring, connectivity, big data, smart assembling, control, and scheduling [47]. AI
in intelligent manufacturing is utilized in various applications such as quality inspec-
tion, energy conservation, supply chain, and predictive maintenance [48]. The lifecycle
of industry 4.0 in smart manufacturing is mentioned below in Fig. 3. While Fig. 4
describes the model of an intelligent manufacturing system and its applications [4].

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 75
Fig. 3. Lifecycle of Industry 4.0 in smart manufacturing [47].
Fig. 4. Intelligent manufacturing model design [4].
Fig. 5. Smart hybrid manufacturing system [2].
Autonomous sensing, learning, analysis, interconnection, cognition, decision-
making, control, and information execution are included in the above figure, which
integrates and optimizes the copious features of a manufacturing enterprise. AI-based

76 A. S. Choudhury et al.
Smart manufacturing leads to enhanced worker safety, product quality, energy use, pro-
duction efficiency, and fault predictions leading to a more high-yielding and secure
workspace, thus engaging smart machines to carry out big tasks and assisting the human
labor force getting rid of routine procedures [5]. Recently, the rise in usage of data-driven
approaches has led to the achievement of monitoring and diagnosis by CPS observation
and analysis that collect and communicate immense data through standardized inter-
faces, which gives rise to the Internet of Things [6]. Figure 5shows an image of an
intelligent hybrid manufacturing system consisting of sensors and new technology such
as big data.
Machinery industries around the world have utilized the technology of smart
machines and the automation of assembly lines to send the production data of these
machines to a monitoring platform in real-time, such as inspection of ball-bearing mal-
function, industrial data-driven monitoring, ball-bearing vibration data, remote wind
turbine condition monitoring, vibration monitoring for smart maintenance and analy-
sis of vibration time [7]. Significant cost reduction is made using predictive mainte-
nance. This method can be proposed by prolonging the functional life of manufacturing
machines and increasing overall equipment effectiveness [8]. In Industry 4.0 manu-
facturing, condition monitoring has been a valuable tool for improving safety, health,
and equipment performance. In smart and sophisticated industrial equipment [9]. The
knowledge-based intelligent supervisory system proposes a pattern recognition strategy
and learning process to inspect rare quality events [10]. In this article, different types of
ML and their applications in Industry 4.0 have been discussed. Also, how AI has taken
place in Indian manufacturing, its scope, possibilities, and suggestions are discussed.
This paper has been written to help future scientists to undergo further research regard-
ing artificial intelligence and its algorithms in smart manufacturing. The article has also
been reported to depict the Indian scenario in Industry 4.0. The Indian Government,
Indian scientists, and engineers will be aware of the actual condition of Indian man-
ufacturing and get encouraged to work with this new-age technology. Simultaneously,
this article will also encourage Indian industrialists to invest capital in India’s AI-based
manufacturing. Last but not least, this manuscript tells us about the consequences of
AI technology implementation in less developed countries, such as unemployment (due
to lack of skill), to aware factory workers as well as skilled professionals of the reality
regarding its actual implementation so that they will become ready to get accustomed
with this new AI-based manufacturing technology.
2 Literature Review
AI and its powerful technologies, such as machine learning (ML), deep learning (DL),
etc., are generally widespread in manufacturing. It has been evident that applying these
technologies in real life requires enormous capital and efficient human resources capable
of cooperative effort in surroundings [1]. The rapid advancement of machine learning has
led to the massive revolution in the artificial intelligence field through which machines
are allowed to learn, improve and optimize specific tasks without being programmed
directly. Machine learning can be used widely in smart machining (consisting of CPSs)
[2]. The Industrial Internet of Things (IIoT) provides real-time production data collecting

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 77
with enhanced wireless connectivity, leading to Industry 4.0 powered by AI [5]. Remark-
able progress has been made in recent years regarding database technologies, computer
power, machine learning (ML), big data, and optimization methods to attain fault-free
(defect-free) processes with the help of ISCS(Intelligent Supervisory Control Systems)
[10]. Predictive model-based quality inspection is an innovative solution developed for
industrial manufacturing applications using edge cloud computing technology, machine
learning techniques, and IOT architecture. The quality inspection processes based on
the predictive model are shown in Fig. 6[11].
Fig. 6. Predictive model-based quality inspection framework arrangement [11].
Automation results in affordable cost, high reliability, and highly endurant qual-
ity inspection process in smart manufacturing industries. This helps to optimize high
productivity, reliability, and repeatability. The automation process in manufacturing is
run and regulated by a Programmable Logic Controller (PLC) [12]. Numerous appli-
cations of data science technologies or big data analytics in industries include process
adjustment, monitoring, and optimization [13]. DL-based fault diagnosis of rotating
machinery eradicates the drawbacks of traditional fault diagnosis methods [14]. Artifi-
cial neural networks (ANNs) have a long history of detecting equipment health conditions
and RUL prediction in smart machines because of their effectiveness, adaptability, and
many other factors [15]. If the engineers accurately implement inactive state detection
in smart appliances in manufacturing, it will be beneficial in performing maintenance
works, error reduction, and catastrophic failure detection [16]. The advancement of the
IIoT has led to the rapid development and installation of sensors to monitor the machine
condition and check whether the machine is operating and working correctly or not [17].
Incorrect readings and values of malfunctioning sensors can be estimated by accurately
performing predictive analysis of big data, which can also be used in decision-making,
including operation and maintenance planning [18]. Artificial intelligence, data mining,
and other applications all use neural networks. A Deep Neural Network (DNN) is mainly
proposed for non-linear high-dimensional regression problems, leading to the ambigu-
ous process due to complexity [19]. Extreme learning methods are generally applied
to eradicate the difficulties of a single hidden layer feedforward network and enhance
generalization performance and learning capability [20]. One of the essential bearing
types is the rolling element bearing. It is commonly used in the mechatronics field. The
various bearing failures of rolling elements affect industrial equipment, such as produc-
tivity reduction, the rise of safety risks, and accuracy loss within this severe and harsh
working environment. RUL (Remaining Useful Life) prediction is helpful in industrial

78 A. S. Choudhury et al.
manufacturing and production optimization [21]. The time-domain vibration signal fea-
tures are extracted through fault diagnosis from the rotating machinery consisting of
standard and flawed bearings. This can be possible with the help of ANN having input,
hidden, and output layers [24]. Figure 7shows various time-domain vibration signals.
Fig. 7. Time-domain vibration signal: (a) acquired (normal), (b) band-pass filtered (normal), (c)
wavelet transformed D2 (normal), (d) acquired (defective), (e) band-pass filtered (defective), (f)
wavelet transformed D2 (normal) [24].
Machine learning techniques like neural networks help maintain and manage the
considerable data complexities [23]. Cyber-Physical Production Systems (CPPS) and
reducing this amount also helps to enhance machine efficiency, leading to more cost-
effective output in industrial plants [22]. Data mining and ML (Machine Learning) algo-
rithms can be executed to the data present in the SAP application to build classification
models for predicting the reliability of industrial machines. [25]. Bearings are a crucial
part of rotating machinery operation in most manufacturing systems. RUL and health
analysis of bearings are performed to predict reliability and safety in manufacturing
by increasingly providing powerful methods and processes that enable smart progno-
sis and bearing health management [26]. Defects present in rolling bearing may result
in machine failure. But to avoid malfunctioning, early detection of faults is essential
[27]. In comparison to other rotating machinery defects, rotor faults (mainly bearing
and gear faults) have attracted more attention from the AI research community regard-
ing the use of fault-specific traits in feature engineering [28]. Therefore, the vibration
analysis technique predicts better reports in rolling bearing condition monitoring and
fault diagnosis [46]. Oil and gas industry projects require colossal capital, including
equipment acquisition and installation. The current drop in petroleum prices has limited
spending, highlighting the necessity of proper maintenance management in the oil and
gas business. Rotating mechanical equipment such as compressors, pumps, and induc-
tion motors are essential elements widely used in manufacturing procedures [28]. Image
analysis powered by artificial intelligence enables accurate material characterization and
measurement, displaying the quality of composite materials [30]. For the high-resolution
images in the dataset, such as (the LCD panel cutting wheel degradation dataset), to
enhance the computational efficiency, one will first extract the regions of interest from
the raw image data, with 1400 ×80 pixels. After that, those regions are transformed into
grey-scale images for further processing and then pretraining unsupervised data based

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 79
on the dataset information [45]. Machine Learning and artificial intelligence will better
identify failures, ensure quality, and improve preventative maintenance in real-world
applications [31].
3 Methodology
We gathered, inspected, and clustered the data relevant to countless websites and dif-
ferent research papers as per the research requirements. The data has been collected
from multiple websites together with the help of a brief introduction to AI technology
and Industry 4.0. After that, the information was put compactly. Different AI and ML
algorithms are used in smart manufacturing [1].
Tabl e 1 . Different models of prediction [11].
Model Accuracy
(%)
Stand ard
deviation
(%)
Recall (%) Pre cision
(%)
Training
time (1000
rows) in ms
Scoring time
(1000 rows)
in ms
Naïve
Bayes
83.5 ±2.7 94.7 75.5 3 9
Decision
Tree
88.2 ±1.5 91.9 84.0 39 6
LR 71.9 ±1.3 77.0 66.8 49 27
SVM 92.9 ±1.3 96.4 89.3 300 360
GBT 92.6 ±1.0 89.9 93.1 240
Methods such as CNN and ELM are applied in gearbox and motor-bearing datasets.
The Continuous Wavelet Transform (CWT) is initially implemented to get pre-processed
presentations of raw vibration signals. After that, the CNN algorithm is developed to
extract high-level features, and ELM is further used to enhance the classification per-
formance [32]. While ANN is used to classify the machine status into standard or faulty
bearings, R-ELM is used to extract stator current vibration signals, detect bearing faults,
and accurately achieve reliable classification, satisfying the need to see online bearing
fault [33,24]. The performance of different prediction models is shown in Table 1.
Various signal processing techniques, such as STFT, WPT, FFT, etc., are proposed to
overcome the challenges, such as removing background noise from vibration signals to
extract the fault features with high resolution [34]. Mainly deep learning algorithms are
used for regression of rotorcraft vibrational spectra [35]. Below at Table 2, it has been
discussed about the input signal effect.
Generative Adversarial Networks (GAN) solve the current problems effectively
encountered in defect examination of industrial datasets and identify unrevealed defects
in future processing events, which led to its increased usage in Industrial Anomaly
Detection [36]. In AI diagnostic techniques, spectral envelope analysis of the current
remnant eliminates noise, manifesting the characteristic bearing faults [37]. Integration

80 A. S. Choudhury et al.
Tabl e 2 . Effects of input signals on identifying machine conditions with five features (RMS, s2,
g3,g4, g6) [24].
Case no Input signals Training success Test success Epochs
1 1 24/24 (100%) 13/16 (81.25%) 28
2 2 24/24 (100%) 14/16 (87.50%) 17
3 3 24/24 (100%) 12/16 (75.00%) 33
4 4 24/24 (100%) 15/16 (93.75%) 24
5 5 24/24 (100%) 15/16 (93.75%) 19
62,3 48/48 (100%) 32/32 (100%) 12
72, 3, 4 72/72 (100%) 48/48(100%) 22
81, 2, 3, 4 96/96 (100%) 63/64 (98.44%) 23
91, 2, 3, 4, 5 120/120 (100%) 79/80 (98.75%) 32
of RNN with LSTM can mitigate risk in rotating equipment predictive maintenance, lead-
ing to cost reduction in oil and gas operations [38]. GDAU Neural Network describes
the tendency of rolling bearing degradation to have more vital short-term and long-
term prediction ability, so it is more worthy for RUL prediction of bearings [21]. After
undergoing extraction from the raw image data, the grey-scale images and pretraining
unsupervised ML-based RUL prediction algorithms such as DCNN, DCNN-M, LSTM,
NoAtt, and Nosupatt are used in the LCD panel cutting wheel degradation dataset con-
taining images of multiple wheels having high-resolution. These RUL prediction meth-
ods provide a practical approach to prognosticative problems and partial observations
[45]. Thus, recently there has been a rise in AI-based predictive maintenance and fault
diagnosis in smart manufacturing, mechanical processes, and machinery.
4 Findings and Discussion
AI is a technology with perceptive and psychological abilities, having some high-yielding
research relevant fields such as image processing, natural language processing, machine
learning, etc., which is currently used in industry 4.0 manufacturing systems. Different
manufacturing abilities such as Computer Numerical Control (CNC), automated guided
vehicles (AGV), Direct Numerical Control (DNC), robotics, etc., are being used in smart
manufacturing. Recently, the Internet of Things has taken manufacturing to another new
level. The disadvantage is that, in many developing and underdeveloped countries such
as India, there is a lack of resources to set up a basic structure; as most businesses
operate in villages, there is a high cost of the smart infrastructure, skills, and training
deficit among people in these technologies and a profitable proper investment put a
barrier to implement this AI-based smart manufacturing technology. In less developed
countries, unemployment is the central issue that led to numerous constraints in the
absolute implementation of artificial intelligence. Other than that, according to experts AI
and new age technologies only become a crisis for people who cannot adapt themselves,

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 81
readjust according to the market’s needs, or fail to become accustomed to new technology
and skills, leading to joblessness. The probability of jobs in various fields due to artificial
intelligence is shown in Fig. 8.
Fig. 8. The perceived probability of jobs due to Artificial Intelligence [45].
5 Research Limitation and Future Scope
Artificial intelligence and machine learning are extensively applied in today’s world in
different fields and purposes. Among them, smart manufacturing is one of the fields
where the implementation of AI technology is at its peak. But, there are many ways
to reap the benefits of artificial intelligence, such as smart maintenance, better product
development, quality improvement, market adaptation, etc. Innovative care means main-
taining manufacturing machines and systems more brilliantly, i.e., reducing the mainte-
nance cost of appliances and types of equipment. As maintenance of equipment is one
of the most significant expenses in manufacturing, it is necessary to implement smart
maintenance such as predictive maintenance (powered by AI algorithms such as neural
networks and machine learning), which will help save enormous amounts of money and
enhance RUL of machinery. Through better product development, one can assess and
examine the different parameters in production, such as available production resources,
budget, and time, which can be implemented with the help of deep learning models and
algorithms. To meet the highest standard and quality of products, machine learning, and
machine vision can be used to identify, detect and eliminate faults in products and alert
about the problems at the production line which may affect the overall production, lead-
ing toward production quality improvement. AI and ML techniques will help the smart
manufacturing industries improve supply chains and strategic vision and make them
interact with changes in the market by generating estimates relating to several factors
like political situation, weather, consumer behavior, the status of the economy, etc. The
utilization of AI, robots, and CPS will probably revolutionize mass production robots.
CPS can perform any laborious tasks at high speed in smart factory units, eradicating
human error and delivering superior levels of quality assurance. Unlike humans, AI and
industrial automation can easily carry out tasks in hazardous places. Overall, AI-run
smart machines can provide skilled workers, engineers, and scientists opportunities to
focus on their complex and innovative functions in science, engineering, and technology
rather than tedious and ordinary human tasks. But, the lack of necessary skills of workers
regarding AI technology, especially in developing countries like India, may hinder the

82 A. S. Choudhury et al.
progress of AI in the Industry 4.0 manufacturing, which can only be solved by educating
and equipping them with these AI-based technologies.
6 Indian Scenario
In India, Industry 4.0 smart manufacturing induces the industrial stalwarts to lay the
groundwork for smart factories and adopt modern and innovative technologies. The
Indian Government has recently initiated the Smart Advanced Manufacturing and Rapid
Transformation Hub (SAMARTH). To improve the application of AI-based smart manu-
facturing in the current context, the Indian Government is developing a National Policy on
Advanced Manufacturing [1]. Our country has achievedtechnological excellence by inte-
grating Cyber-Physical Systems (CPS) and Information and Communication Technolo-
gies (ICT) into Advanced Manufacturing Technologies (AMTs). Increased automation
in additive manufacturing, Advanced Manufacturing Systems, manufacturing robotics,
advanced analytics, and Big Data are all worth mentioning in the design of smart manu-
facturing for Industry 4.0. They will help Indian MSMEs become more internationally
competitive and contribute to global value creation [39]. Though adoption of artifi-
cial intelligence is less in India, there has been a remarkable transformation in all the
Indian industrial sectors where companies are adopting, developing, and integrating AI
technologies in their products and industrial processes, such as electronics, heavy elec-
tricals, automobiles, fintech, software/IT, agriculture, agrobased industries, etc. [40]. In
terms of government funding, the Union Cabinet approved the launch of the National
Mission regarding Interdisciplinary Cyber-Physical Systems (NM-ICPS) in 2019, which
the Department of Science and Technology (DST) will execute with an unlimited budget
of INR 3660 Cr (USD 494 Mn) for five years to make India a leader in Cyber-Physical
Systems (which includes AI, ML, and IoT) (FY 2019–20 to 2023–24). The mission’s
goal is to build a strong and stable ecosystem for CPS technologies in India, which
would help the country’s Industry 4.0 manufacturing sector thrive [41]. SMEs have sig-
nificant advantages in terms of innovation in general, but they face a variety of obstacles
in India [42]. Though several countries have decided on their strategy for AI, India has
not yet formulated its strategy in Industry 4.0 [43]. Another disadvantage is the lack of
skilled workers in AI technology in our country; unemployment will rise in India. On
one side cities will be equipped with all modern facilities and will be becoming smart
and other side jobs will be killed due to transformation. Low and Middle skills level
jobs will be shrunk, but high-skilled jobs where the critical decision will have to take
will exist as machines cannot resemble human intelligence in case of making critical
decisions. This transformation will add new development aspects to India’s infrastruc-
ture and enhance the economic status in the coming years. However, few jobs in a few
sectors will disappear because of transformation through AI in the next 5 to 10 years
[44].
7 Conclusion
Although AI is still considered a nascent stage in Industry 4.0 manufacturing, one can still
hopefully say that technological transformations are occurring. 5G technologies in com-
munication can improve the overall efficiency and productivity, which has high network

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 83
reliability and support IoT and CPS devices according to the industry requirements.
Other advancements include lights-out manufacturing, which can create and regulate
production with minimal human interaction, and smart and dynamic technology, which
can be effective in areas with high production rates and low human error rates. Setting
up an AI infrastructure platform may be costly due to advanced machines and equip-
ment, but this reduces the labor required to finish the final product. But the advanced
technology of AI-based applications in the Indian scenario will be extracted fully in the
SME sector, which can be achieved 100% by providing incentives and encouragement
to SMEs (because most of the people in India are employed in this sector). Similarly,
the Indian educational system needs to be enhanced to enormously extract the potential
benefits of these technologies.
References
1. Rizvi, A.T., Haleem, A., Bahl, S., Javaid, M.: Artificial intelligence (AI) and its applications
in indian manufacturing: a review. In: Acharya, S.K., Mishra, D.P. (eds.) Current Advances in
Mechanical Engineering. LNME, pp. 825–835. Springer, Singapore (2021). https://doi.org/
10.1007/978-981-33-4795-3_76
2. Kim, D.-H., et al.: Smart machining process using machine learning: a review and perspective
on machining industry. Int. J. Precis. Eng. Manuf. Green Technol. 5(4), 555–568 (2018).
https://doi.org/10.1007/s40684-018-0057-y
3. Trakadas, P., et al.: An Artificial intelligence-based collaboration approach in industrial IoT
manufacturing: key concepts, architectural extensions and potential applications. Sensors
20(19), 5480 (2020). https://doi.org/10.3390/s20195480
4. Li, B., Hou, B., Yu, W., Lu, X., Yang, C.: Applications of artificial intelligence in intelligent
manufacturing: a review. Front. Inf. Technol. Electron. Eng. 18(1), 86–96 (2017). https://doi.
org/10.1631/FITEE.1601885
5. Angelopoulos, A., et al.: Tackling faults in the industry 4.0 era—a survey of machine-learning
solutions and key aspects. Sensors 20(1), 109 (2020). https://doi.org/10.3390/s20010109
6. Kumar, M., Aggarwal, A., Rawat, T.K.: Bat algorithm: application to adaptive infinite impulse
response system identification. Arab. J. Sci. Eng. 41(9), 3587–3604 (2016)
7. Tsai, M.-F., Chu, Y.-C., Li, M.-H., Chen, L.-W.: Smart machinery monitoring system with
reduced information transmission and fault prediction methods using industrial Internet of
Things. Mathematics 9(1), 3 (2021). https://doi.org/10.3390/math9010003
8. McCulloch, W.S., Pitts, W.: A logical calculus of the ideas immanent in nervous activity. Bull.
Math. Biophys. 5(4), 115–133 (1943). https://doi.org/10.1007/BF02478259
9. Hotait, H., Chiementin, X., Rasolofondraibe, L.: Intelligent online monitoring of rolling
bearing: diagnosis and prognosis. Entropy 23(7), 791(2021)
10. Escobar, C.A., Morales-Menendez, R.: Machine learning techniques for quality control in
high conformance manufacturing environment. Adv. Mech. Eng. 10(2), 1–16 (2018). https://
doi.org/10.1177/1687814018755519
11. Pai, P.F., Hong, W.C.: Forecasting regional electricity load based on recurrent support vector
machines with genetic algorithms. Electric Power Syst. Res. 74 (3), 417–425 (2005)
12. Ashwini, K., Rudraswamy, S.B.: Automated inspection system for automobile bearing seals.
Mater. Today Proc. 46(10), 4709–5471 (2020). https://doi.org/10.1016/j.matpr.2020.10.301
13. Butte, S., Prashanth, A.R., Patil, S.: Machine learning based predictive maintenance strat-
egy: a super learning approach with deep neural networks. In: 2018 IEEE Workshop on
Microelectronics and Electron Devices (WMED), pp. 1–5 (2018)

84 A. S. Choudhury et al.
14. Tang, S., Yuan, S., Zhu, Y.: Deep learning-based intelligent fault diagnosis methods toward
rotating machinery. IEEE Access 8, 9335–9346 (2020)
15. Tian, Z.: An artificial neural network method for remaining useful life prediction of equipment
subject to condition monitoring. J. Intell. Manuf. 23(1), 227–237 (2012). https://doi.org/10.
1007/s10845-009-0356-9
16. Borith, T., Bakhit, S., Nasridinov, A., Yoo, K.-H.: Prediction of machine inactivation status
using statistical feature extraction and machine learning. Appl. Sci. 10(21), 7413 (2020).
https://doi.org/10.3390/app10217413
17. Ertu˘grul, Ö.F.: A novel approach for extracting ideal exemplars by clustering for massive
time-ordered datasets. Turk. J. Electr. Eng. Comput. Sci. 25(4), 2614–2634 (2017). https://
doi.org/10.3906/elk-1602-341
18. Miorandi, D., Sicari, S., De Pellegrini, F.: Internet of things: vision, applications and research
challenges. Ad Hoc Netw. 10(7), 1497–1516 (2012). https://doi.org/10.1016/j.adhoc.2012.
02.016
19. Beyerer, J., Usländer, T.: Industrial internet of things supporting factory automation.
at-Automatisierungstechnik 64(9), 697–698 (2016). https://doi.org/10.1515/auto-2016-0104
20. Ding, S., Zhao, H., Zhang, Y., Xu, X., Nie, R.: Extreme learning machine: algorithm, theory
and applications. Artif. Intell. Rev. 44(1), 103–115 (2013). https://doi.org/10.1007/s10462-
013-9405-z
21. Qin, Y., Chen, D., Xiang, S., Zhu, C.: Gated dual attention unit neural networks for remaining
useful life prediction of rolling bearings. IEEE Trans. Ind. Inf. 17(9), 6438–6447 (2021).
https://doi.org/10.1109/TII.2020.2999442
22. Kroll, B., Schaffranek, D., Schriegel, S., Niggemann, O.: System modeling based on machine
learning for anomaly detection and predictive maintenance in industrial plants. In: Proceedings
of the 2014 IEEE ETFA, pp. 1–7 (2014). https://doi.org/10.1109/ETFA.2014.7005202
23. Dubois, D., Prade, H.: Possibility theory is not fully compositional! A comment on a short
note by H.J. Greenberg. Fuzzy Sets Syst. 95(1), 131–134 (1998)
24. Krishnasamy, L., Khan, F., Haddara, M.: Development of a risk-based maintenance (RBM)
strategy for a power-generating plant. J. Loss Prev. Process Ind. 18(2), 69–81 (2005). https://
doi.org/10.1016/j.jlp.2005.01.002
25. Shilaskar, S., Ghatol, A., Chatur, P.: Medical decision support system for extremely
imbalanced datasets. Inf. Sci. 384, 205–19 (2017). https://doi.org/10.1016/j.ins.2016.08.077
26. Rena, L., Suna, Y., Cuia, J., Zhang, L.: Bearing remaining useful life prediction based on deep
autoencoder and deep neural networks. J. Manuf. Syst. 48(C), 71–77 (2018). https://doi.org/
10.1016/j.jmsy.2018.04.008
27. Gupta, P., Pradhan, M.K.: Fault detection analysis in rolling element bearing: a review. Mater.
Today Proc. 4(2), 2085–2094 (2017)
28. Nath, A.G., Udmale, S.S., Singh, S.K.: Role of artificial intelligence in rotor fault diagnosis:
a comprehensive review. Artif. Intell. Rev. 54(4), 2609–2668 (2020). https://doi.org/10.1007/
s10462-020-09910-w
29. Golub, T.R., et al.: Molecular classification of cancer: class discovery and class prediction
by gene expression monitoring. Science 286(5439), 531–7(1999). https://doi.org/10.1126/sci
ence.286.5439.531
30. Aggour, K.S., et al.: Artificial intelligence/machine learning in manufacturing and inspection:
a GE perspective. MRS Bull. 44(7), 545–558 (2019). https://doi.org/10.1557/mrs.2019.157
31. Mohapatra, P., Chakravarty, S., Dash, P.K.: Microarray medical data classification using kernel
ridge regression and modified cat swarm optimization based gene selection system. Swarm
Evolut. Comput. 28, 144–60 (2016). https://doi.org/10.1016/j.swevo.2016.02.002
32. Chen, Z., Gryllias, K., Li, W.: Mechanical fault diagnosis using convolutional neural networks
and extreme learning machine. Mech. Syst. Signal Process. 133(1), 106272 (2019). https://
doi.org/10.1016/j.ymssp.2019.106272

Implementation of Artificial Intelligence (AI) in Smart Manufacturing 85
33. Zhang, H.-G., Zhang, S., Yin, Y.-X.: A novel improved ELM algorithm for a real industrial
application. Math. Probl. Eng. 2, 1–7 (2014). https://doi.org/10.1155/2014/824765
34. García-Nieto, J., Alba, E.: Parallel multi-swarm optimizer for gene selection in DNA
microarrays. Appl. Intell. 37(2), 255–266 (2012). https://doi.org/10.1007/s10489-011-0325-9
35. Martinez, D., Brewer, W., Behm, G., Strelzoff, A., Wilson, A., Wade, D.: Deep learning
evolutionary optimization for regression of rotorcraft vibrational spectra. In: 2018 IEEE/ACM
Machine Learning in HPC Environments (MLHPC), pp. 57–66 (2018). https://doi.org/10.
1109/MLHPC.2018.8638645
36. Wang, A., An, N., Chen, G., Yang, J., Li, L., et al.: Incremental wrapper based gene selec-
tion with Markov blanket. In: 2014 IEEE International Conference on Bioinformatics and
Biomedicine (BIBM), pp. 74–79. IEEE (2014). https://doi.org/10.1109/BIBM.2014.6999251
37. Bolón-Canedo, V., Sánchez-Maroño, N., Alonso-Betanzos, A.: Distributed feature selection:
an application to microarray data classification. Appl. Soft Comput. 30, 136–150 (2015).
https://doi.org/10.1016/j.asoc.2015.01.035
38. Chazhoor,A., Mounika, Y., Vergin Raja Sarobin, M., Sanjana, M.V., Yasashvini, R.: Predictive
maintenance using machine learning based classification models. IOP Conf. Ser. Mater. Sci.
Eng. 954(1) (2020). https://doi.org/10.1088/1757899X/954/1/012001
39. Sankararaju, M., Dharmar, S.: Design of low power CMOS LC VCO for direct conversion
transceiver. Turk. J. Electr. Eng. Comput. Sci. 24(4), 3263–3273 (2016)
40. Hossain, M., Muhammad, G., Guizani, N.: Explainable AI and mass surveillance system-
based healthcare framework to combat COVID-19 like pandemics. IEEE Network 34(4),
126–132 (2020). https://doi.org/10.1109/MNET.011.2000458
41. Acar, E., Yilmaz, I.: COVID-19 detection on IBM quantum computer with classical-quantum
transfer learning. Turk. J. Electr. Eng. Comput. Sci. 29(1), 46–61 (2021). https://doi.org/10.
3906/elk-2006-94
42. Krishnaswamy, K.N., Bala Subrahmanya, M.H., Mathirajan, M.: Technological innovation
induced growth of engineering industry SMEs: case studies in Bangalore. Asian J. Innov.
Policy 4(2), 217–41(2015). https://doi.org/10.7545/AJIP.2015.4.2.217
43. Conti, M, Dehghantanha, A., Franke, K., Watson, S.: Internet of things security and forensics:
challenges and opportunities. Future Gener. Comput. Syst. 78(2), 544–546 (2018). https://
doi.org/10.1016/j.future.2017.07.060
44. Mendoza, C.V., Kleinschmidt, J.H.: Mitigating on-off attacks in the Internet of Things using
a distributed trust management scheme. Int. J. Distrib. Sens. Netw. 11(11), 859731 (2015)
45. Chen, R., Guo, J., Bao, F.: Trust management for SOA-based IoT and its application to service
composition. IEEE Trans. Serv. Comput. 9(3), 482–95 (2014)
46. Abderrahim, O.B., Elhedhili, M.H., Saidane, L.: DTMS-IoT: a Dirichlet-based trust man-
agement system mitigating OnOff attacks and dishonest recommendations for the Internet of
Things. In: IEEE/ACS 13th International Conference of Computer Systems and Applications
(AICCSA), Agadir, Morocco, pp. 1–8 (2016)
47. Zheng, P., et al.: Smart manufacturing systems for Industry 4.0: conceptual framework, sce-
narios, and future perspectives. Front. Mech. Eng. 13(2), 137–150 (2018). https://doi.org/10.
1007/s11465-018-0499-5
48. Ding, H., Gao, R.X., Isaksson, A.J., Landers, R.G., Parisini, T., Yuan, Y.: State of AIBased
monitoring in smart manufacturing and introduction to focused section. IEEE/ASME Trans.
Mechatron. 25(5), 2143–2154 (2020). https://doi.org/10.1109/TMECH.2020.3022983

Emerging Computing Computational
Intelligence

Flight Fare Prediction Using Machine Learning
K. P. Arjun1(B), Tushar Rawat2, Rohan Singh2, and N. M. Sreenarayanan1
1Department of Computer Science and Engineering, GITAM University, Bengaluru, Karnataka,
India
arjunkppc@gmail.com, sree.narayanan1@gmail.com
2School of Computer Science and Engineering, Galgotias University, Greater Noida, Uttar
Pradesh, India
tusharrawat517@gmail.com, rohansingh7217@gmail.com
Abstract. The price of airline tickets can fluctuate gradually and generally with
the same aircraft, independent of, in the seats that are closest together inside
the same cabinet. Customers have the expectation that they will pay decreased
expenses, whereas airlines work to maintain or even increase their overall earnings
while also working to improve their profitability. To maximise their payload,
airlines employ a variety of mathematical methods, such as guessing and suitable
classification, among others. Models that estimate the best open door buy and
models that anticipate the cost of a basic ticket are the two sorts of client-side
models that various industry professionals recommend in order to save clients
money. Both of these models fall under the category of client-side predictive
models. According to our research, models on both sides depend on the restricted
performance of several components, such as actual ticket price data, the date
the ticket was purchased, and the date the passenger exited the venue. Many
individuals take flights on a regular basis, and as a result, they are familiar with
the times of year that provide the best deals on airline tickets. Despite this, there
are a great number of individuals who have recently purchased tickets but wind
up falling prey to the snares made by organisations, as a result of which they wind
up spending more money than they should have.
Keywords: Airline price ·Machine learning ·Flutter ·Flask ·Random Forest ·
Flight ticket
1 Introduction
Calculations of forecasts are essential to the process of matching customers with the
appropriate products and typically involve the use of real customer data. A related con-
dition does not strictly anticipate the future item but rather suggests an item that does
not occur in the actual information but that the customer could appreciate. In most cases,
the focus of these proposal methods is positioned appropriately at the appropriate point
in a sequence [1].
Despite this, there are situations in which it is necessary to predict as well as maybe
suggest. A good illustration of this is the process of booking flights, in which the objective
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 89–99, 2022.
https://doi.org/10.1007/978-3-031-22915-2_8

90 K. P. Arjun et al.
examples for the guaranteed customer might either be constant or change depending on
a huge number of complex elements [2].
This challenging combat booking area is the focal point of this study, with a spe-
cific interest in measuring client ways of behaving for this mixed forecast proposal
environment. Specifically, this exploration has a specific interest in gauging client ways
of behaving. Researching the efficacy of algorithmic ways in suggesting the follow-
ing objective reserving for a carrier client is the purpose of our examination. The vast
majority of the previous work in both the expectation and suggestion sectors has been
developed and evaluated solely on verifiable datasets. This is true of both regions. A lim-
ited number of earlier examinations have conducted research with actual customers to
evaluate the generated models. In the course of this investigation, we conduct evaluations
using both genuine information and actual customers [3].
The purpose of this effort is to promote the development of an application that will
use the AI model to predict the cost of flight for a variety of airlines. The customer will
receive the features that were anticipated, and using it as a reference, the customer may
choose to purchase his tickets in the same manner. As a result of the same problem,
airlines are attempting to keep a tighter rein on the prices of their tickets in order to
boost their earnings [4]. Numerous individuals have made flying their primary mode of
transportation, and as a result, they are constantly on the lookout for ways to cut costs
when they make their reservations. However, there are a lot of people who are not used
to purchasing tickets, and as a result, they frequently find themselves falling into the
wrong trap set by organisations, which results in them paying more money than they
should have. The proposed structure has the potential to assist clients save a significant
amount of rupees by displaying booking information to them via the most advantageous
open door [5]. We have constructed a model out of wood, however it is not appropriate
for use in estimating the cost of aircraft due to the many different aspects that play a role
in determining the cost of aircraft [6].
To estimate future trip costs, our team came up with the Random Forest Regression
Algorithm as a result of the fact that it employs both regression and classification in its
prediction-making process, resulting in a more precise outcome. The research that we
carried out lends credence to this notion [7].
The cost of airline tickets can be hard to see, today we can see the price, see the
cost of the same flight tomorrow will be another matter. We may have often heard
travelers say that the cost of airline tickets is unpredictable [8]. Air travel has become
an important means of transportation travel a significant distance. To maximize profits,
airlines use an integrated pricing system called “yield the board” to calculate the cost
of each trip. Competition, etc. The ultimate goal of earning extra profit on each aircraft.
Since travelers are generally willing to admit that air travel costs are increasing when
the purchase date is closer to the departure date, they usually purchase airline tickets
from a city farther away from the departure date as expected [9]. However, buying this
way is not right. It is like failing and in the process the passenger will be spending a lot
of money.
After modeling the guessing system, it was important to make a visual interface that
is easy to use and can be used on any device running on any operating system. AI is
a mathematical investigation that will work best through experience [10]. Officially he

Flight Fare Prediction Using Machine Learning 91
works at work, based on performance, due to involvement. It is a topic under Artifi-
cial Intelligence. While AI controls the in-depth functions performed by a non-human
professional, ML controls selective knowledge-based choices. AI is a very large field
in software engineering. AI can be redirected, or on the other hand can be redirected.
Problems in ML included. Compilation: if a few details are given, we need to plan a
specific way to put it together. Repetition: given a small amount of information, can we
expect the results to be respected? The information used in ML can be of three types:
Categorical Nominal, Categorical Ordinal, and Continuous. We look forward to working
with duplicate models that we suspect can provide us with more accurate results. Filtered
models to work with [11].
In fact, it is undoubtedly a challenge for travelers to foresee when the best opportunity
to buy war tickets is for the following reasons:
•Incomplete Information: Travelers can access part of the network carrier data. Truth
be told, they do not have access to important information, such as the number of extra
tickets and understanding between network company organizations [12].
•Different Information: Data that can be obtained by sailors is categorized. For example,
it is undoubtedly a challenge for the average inspector to find a relationship between
flight costs and flight costs, such as travel expenses, departure time, and so on.
•Unusual Changes: Although inspectors cannot collect a guaranteed flight amount, the
cost change is not smooth. In fact, not all reports are predictable. So, travelers can
only anticipate future flight costs with great effort in terms of recorded prices.
2 Literature Review
A dataset consisting of 1814 information trips on Aegean Airlines was gathered and
used to prepare an artificial intelligence model for the research work. This procedure
was used for the research work that was proposed by K. Tziridis T [13] on air fare cost
expectation using machine learning procedure. To demonstrate how the identification of
highlights can affect the accuracy of a model, a varied number of elements was used to
prepare each model.
The next piece of research to be provided is a concentration on suggestion that
William Forests, an expert, should work on improving buy timing for customers. In
order to construct a model, an incomplete least square relapse technique is applied.
A study on aeroplane passage anticipation using AI calculation uses a small dataset
consisting of travels between Delhi and Bombay, as stated by the author Supriya Rajankar
[14]. The calculations K-closest neighbours (KNN), straight relapse, and support vector
machine (SVM), among others, are applied.
Research carried out by Santos [15] investigates the cost of flying from Madrid to
London, Frankfurt, New York, and Paris over the course of a few short months. The
number of days that are considered to be adequate before booking an airline ticket is
provided by the model. Tianyi Wang [16] developed a framework in which two informa-
tion bases are integrated along with information regarding macroeconomics and artificial
intelligence calculations, such as support vector machine.
The aforementioned algorithms each have their own drawbacks, such as the fact that
there is not enough data in the system to make an accurate prediction. The accuracy

92 K. P. Arjun et al.
of the system shifts whenever the algorithm is altered, which can make things a little
bit confused, despite the fact that the accuracy shifts significantly only when essential
elements are disabled. These studies are now considered to be obsolete as a result of the
proliferation of new airlines, significant shifts in the cost of oil, and rising prices for a
variety of other goods and services [17–20].
3 Proposed Methodology
In order to predict the price of a plane it was necessary to consider all possible parameters
and how it effects the price of the aircraft in order to improve the Random Forest machine
model which provides almost the most accurate result based on the data provided. Table 1
represents the columns name and its description.
Tabl e 1 . Parameters for price prediction.
Name Description
Origin The place where flight will star
from
Destination Place where the flight has to
reach
Departure date The departure date of the flight
Arrival date The arrival date of the flight
Departure time The departure time of the flight
“HH:MM”
Arrival time The arrival time of the flight
“HH:MM”
Airline company The airline company whose
flight
we are using
Duration Total time taken by the flight to
complete the journey
Stops Number of stops between origin
and destination
To make a for “Airfare Prediction” model in light of past carrier ticket deals dataset for
further developing deals in Indian Domestic Airline. Our fundamental thought process
is to furnish the client with a forecast framework from which it can take an ideal choice
of expanding or diminishing the Airfare so the flight doesn’t go unfilled or no cash is
lost because of unexpected expansion in unrefined petroleum [8].
a) To perform information investigation on client’s ticket booking information for a
short measure of time.

Flight Fare Prediction Using Machine Learning 93
b) To refine the information for example Eliminating copy records, vagueness and so
forth.
c) To perform Feature designing to separate significant component from dataset for
expectation.
d) To Brainstorm the Features for example to choose how to utilize those elements
e) To make highlights for example to get new highlights from those helpful elements.
The proposed framework is made out of four stages [13]:
1. Dataset Selection
2. Data Cleaning
3. Feature Extraction
4. Machine Learning Model Selection
3.1 Information Input
Input data is given to the system in the form of a.csv file. The dataset that was chosen from
Kaggle serves as both the training dataset and the testing dataset. The data only pertain
to flights within the country. In total, our dataset is comprised of 11 columns. [https://
www.kaggle.com/datasets/shubhambathwal/flight-price-prediction]. Figure 1shows the
columns header listing.
Fig. 1. Columns in training dataset

94 K. P. Arjun et al.
3.2 Data Cleaning
The process of cleaning data involves removing any instances of null values from the
dataset and replacing them with more appropriate values. These values are typically the
mean, median, or mode of the other data in the column. The presence of null values in
the dataset may have an impact on the accuracy of the model. The data cleaning steps
shows in Fig. 2.
Fig. 2. Cleaning of data
3.3 Feature Extraction
In this phase we try to extract new features from the dataset which will help to train
model more accurate and prediction becomes easy and convenient shows in Fig. 3.New
Features is added to the dataset which becomes the discriminating factor of price of flight
and the reason of their variation. Figure 4shows the correlation between the different
feature in the dataset.
Features that can be considered as deciding factor of flight fare are.
•Feature 1: date and time of time
•Feature 2: date and time of departure
•Feature 3: How the early the ticket is booked
•Feature 4: Type of passenger (Adult/Child)
•Feature 4: Class of the flight booked (Economy/Business)
•Feature 5: Departure Location
•Feature 6: Destination Location

Flight Fare Prediction Using Machine Learning 95
Fig. 3. Feature extraction in model.
Fig. 4. Correlation between attributes
3.4 Machine Learning Model Selection
There are bunch of Machine Learning algorithm to choose from each having their own
pros and cons. Linear regression being easy to train and simple to test but with less
accuracy we decide not to move forward with it. Decision trees are essentially of two
kinds of arrangement and regression tree where arrangement is utilized for unmitigated
values and regression is utilized for persistent qualities. Decision tree picks autonomous
variable from dataset as choice hubs for independent direction.
Random forest fundamentally utilizes gathering of decision trees as gathering of
models. Random amount of information is passed to choice trees and every decision tree
predicts values as indicated by the dataset given to it. From the expectations went with
by the decision trees the typical worth of the anticipated qualities whenever considered

96 K. P. Arjun et al.
as the result of the arbitrary woods model. Since it uses both regression and classification
we find it to be the best fit for our system.
4 Result and Discussions
In proposed work, developed various algorithms such as Linear Regression, Decision
Tree Decision, Random Forest Depression and compared the accuracy of the results
based on our set of experimental data. Based on various levels of accuracy we find that
the Random Forest Regression provides the highest accuracy at 81% shows in Table 2.
So we selected Random Forest Resolve and built user interface based on it.
Tabl e 2 . Accuracy of different algorithms
Algorithms Accuracy
Linear Regression 0.61
Decision Tree
Regression
0.64
Random Forest
Regression
0.85
4.1 Performance Metrics
Performance measurements are validated models that will be used to determine the
accuracy of AI models suitable for various calculations. The sklearn.metrics module will
be used to apply the deficiencies in each model using backslide scales. The following
measurements will be used to assess the bumble level of each model.
4.1.1 MAE (Mean Absolute Error)
A small component of mathematical accuracy is called the Mean Absolute Error (MAE)
as Eq. 1. Mean Absolute Error is basically, as the name suggests, a description of obvious
errors. Direct error is the actual value of the difference between the expected value and
the actual value. It Means Perfect Error It means measuring accuracy of a fixed object.
MAE =1
nx−x

(1)
n=the number of errors,
=summation symbol (which means “add them all up”),
x−ˆx=the absolute errors.
Lesser the value of MAE the better the performance of your model.

Flight Fare Prediction Using Machine Learning 97
4.1.2 MSE (Mean Square Error)
Mean Square Error squares the distinction of real and anticipated result esteems prior to
adding them all rather than utilizing the outright worth shows in Eq. 2.
MSE =1
n∗(actual −forecast)(2)
n=number of items,
=summation notation,
Actual =original or observed y-value,
Forecast =y-value from regression.
4.1.3 RMSE
It is more noticeable than MAE and lower RMSE value among various models to improve
the presentation of that model shows in Eq. 3. R2 (Coefficient of assurance) Helps you
to see how the free factor has changed with the flexibility of your model.
RMSE =(Forecast −Actual)2
n(3)
To use the random tree regression, we used a number of scales like 1000 and the
number of random circuits was 42. This measurement process is well suited for informal
data where dependence between factors is difficult to identify. The Fig. 5shows the
proposed random forest method’s performance matrix.
Fig. 5. Metrics of Random Forest algorithm
5 Conclusion and Future Scope
At the moment, there are a great deal of domains in which management is predicated on
expectancies. One such domain is stock trading and management, which uses items that
reflect the number of shares traded, such as Zestimate, which gives a proven quantity of
the costs associated with real estate. In the airline industry, a need for management like
this that can assist customers in booking tickets has arisen as a direct result of this need.

98 K. P. Arjun et al.
There has been a significant amount of study conducted on this topic making use of a
variety of methods, and additional testing is anticipated to work towards understanding
expectations through the use of a variety of statistics. Information that is more accurate
and has better features can be used in the same way to get results that are more accurate.
In future, our research could be expanded to include air exchange ticketing data,
which could provide additional insight into a specific schedule, such as time and date of
departure, appearance, coverage, etc. Model weather forecast for daily flight or hourly
rate. In addition, the cost of a flight on the market segment may be affected by the
unpredictable influx of large numbers of travelers brought about by different events.
References
1. Sharma, L., Carpenter, M. (eds.): Analysis of machine learning techniques for airfare predic-
tion. In: Computer Vision and Internet of Things: Technologies and Applications, 1st edn.
Chapman and Hall/CRC (2022). https://doi.org/10.1201/9781003244165
2. Khandelwal, K., Sawarkar, A., Hira, S.: A novel approach for fare prediction using machine
learning techniques. Int. J. Next Gener. Comput. Suppl. 12(5), 602–609 (2021). 8p.
3. Arjun, K.P., Achuthshankar, A., Soumya, M.K., Sreenarayanan, N.M., Priya, V.V., Faby, K.A.:
PROvacy: protecting image privacy in social networking sites using reversible data hiding.
In: 2016 10th International Conference on Intelligent Systems and Control (ISCO), pp. 1–4,
January 2016
4. Gupta, J., Singh, I., Arjun, K.P.: Artificial Intelligence for Blockchain I, Blockchain, Internet
of Things, and Artificial Intelligence, vol. 6. CRC Press, April 2021
5. Achuthshankar,A., Achuthshankar, A., Arjun, K., Sreenarayanan, N.: Encryption of reversible
data hiding for better visibility and high security. Procedia Technol. 25, 216–223 (2016)
6. Groves, W., Gini, M.: An agent for optimizing airline ticket purchasing. In: Ito, Jonker, Gini,
Shehory (eds.) Proceedings of the 12th International Conference on Autonomous Agents and
Multiagent Systems (AAMAS 2013), Saint Paul, Minnesota, USA, 6–10 May 2013
7. Biswas, P., et al.: Flight price prediction: a case study. Int. J. Res. Appl. Sci. Eng. Tech-
nol. (IJRASET) 10(6) (2022). https://doi.org/10.22214/ijraset.2022.43666. ISSN: 2321-9653
8. Champawat, J.S., Arora, U., Vijaya, K.: Indian flight fare prediction: a proposal. Int. J. Adv.
Technol. Eng. Sci. 9(3) (2021)
9. Tian, H., Presa-Reyes, M., Tao, Y., et al.: Data analytics for air travel data: a survey and new
perspectives. ACM Comput. Surv. 54(8), 1–35 (2022)
10. Joseph, J., et al.: Flight ticket price predicting with the use of machine learning. Int. J. Adv.
Trends Comput. Sci. Eng. 10(2), 1243–1246 (2021). https://doi.org/10.30534/IJATCSE/2021/
1071022021
11. Oza, R.K., Jain, A.V., Raval, A.S.: Machine learning techniques for predicting airfare prices
based on reviews 9(3) (2020). ISSN: 2319-8753 ISSN: 2347-6710
12. Tziridis, K., Kalampokas, T., Papakostas, G., Diamantaras, K.: Airfare price prediction using
machine learning techniques. In: 25th European Signal Processing Conference (EUSIPCO)
(2017)
13. Rajankar, S., Sakhrakar, N., Rajankar, O.: Flight fare prediction using machine learning
algorithms. Int. J. Eng. Res. Technol. (IJERT) (2019)
14. Santos Domínguez-Menchero, J., Rivera, J., Torres-Manzanera, E.: Optimal purchase timing
in the airline market. J. Air Transp. Manag. 40, 137–143 (2014). ISSN 0969-6997
15. Shabana, T., Afifa, S., Naziya, S., Mariya, K.: A novel machine learning methodology to
increase sales in business services. Int. J. Comput. Sci. Eng. 6(12), 924–926 (2018)

Flight Fare Prediction Using Machine Learning 99
16. Wang, T., et al.: A framework for airfare price prediction: a machine learning approach. In:
2019 IEEE 20th International Conference on Information Reuse and Integration for Data
Science (IRI), pp. 200–207 (2019). https://doi.org/10.1109/IRI.2019.00041
17. Thirumuruganathan, S., Jung, S., Robillos, D.R., Salminen, J., Jansen, B.J.: Forecasting the
nearly unforecastable: why aren’t airline bookings adhering to the prediction algorithm?
Electron. Commer. Res. 21(1), 73–100 (2021)
18. Ratnakanth, G.: Prediction of flight fare using deep learning techniques. In: International
Conference on Computing, Communication and Power Technology (IC3P), pp. 308–313
(2022). https://doi.org/10.1109/IC3P52835.2022.00071
19. Subramanian, R.R., Murali, M.S., Deepak, B., Deepak, P., Reddy, H.N., Sudharsan, R.R.:
Airline fare prediction using machine learning algorithms. In: 2022 4th International Con-
ference on Smart Systems and Inventive Technology (ICSSIT), pp. 877–884 (2022).: https://
doi.org/10.1109/ICSSIT53264.2022.9716563
20. Abdella, J.A., Zaki, N., Shuaib, K., Khan, F.: Airline ticket price and demand prediction: a
survey. J. King Saud Univ. Comput. Inf. Sci. 33(4), 375–391 (2021)

Impact of Work from Home During Covid-19
on the Socio-economic Status of India
Poonam Ojha1, Sudhanshu Maurya2(B), and Manish Kumar Ojha3
1School of Management, Graphic Era Hill University Bhimtal Campus, Nainital 263132,
Uttarakhand, India
2School of Computing, Graphic Era Hill University Bhimtal Campus, Nainital,
Uttarakhand 263132, India
dr.sm0302@gmail.com
3Amity University Noida, Noida, Uttar Pradesh, India
Abstract. Socioeconomic status (SES) is an instrument to measure the economic
and social status of an individual or an economy concerning others. Though,
Socioeconomic status is more commonly used to represent an economic differ-
ence in any society. Work from home is now a day (Covid-19) contributing to
the nation for its socio-economic activities. This paper has examined the impact
of ‘work from home’ on the socio-economic status of India as so many people
became unemployed, the income of the society decreased as well as the Education
system was worse affected. The present situation of the pestilence provided great
importance to work from home (WFH) for many employees to have the opportu-
nity to both carries on working and safely from the risk of virus vulnerability. As
this Pandemic period is uncertain, working from home is more acceptable as the
new normal working way. On the contrary, to find the impact of WFH on socioe-
conomic status, we took three variables: education, employment, and income &
wealth.
Keywords: Work from home ·Socio-economic status ·Education ·Income &
wealth ·Employment
1 Introduction
The socioeconomic study refers to the interaction between the social and economic
behavior of a group of people, linking financial and social issues together. SES is a
prominent indicator of any nation’s economic as well as social position in the world.
This index decides the togetherness of socio-economic activities. “Pandemics are not
a new experience for the communities as they were recorded since prehistoric times.
During each pandemic, major changes were noticed in the areas of economics, local
and national policies, social behavior, and citizens’ mentalities as well. Opposing these
changes, it was detected that mentalities and social behavior were slightest potted as
the institutionalized modifications [1], through public policies, were not adequately
attached and synthesized with the psychosocial changes [2].” During the Pestilence of
Covid-19, it is realized that SES has been affected severely because of aberration of
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
R. Mehra et al. (Eds.): ICCISC 2022, CCIS 1672, pp. 100–113, 2022.
https://doi.org/10.1007/978-3-031-22915-2_9

Impact of Work from Home During Covid-19 101
social and economic activities. The COVID-19 pandemic is becoming furious and will
have its long-term effects worldwide, most probably resulting in structural effects on the
socio-economic status of India and other affected countries. “Like any other epidemics,
COVID-19 has caused noteworthy changes on all levels of modern-day society [3–8].”
The countrywide lockdown has ended up with financial losses as well as affected all
segments of society including health, healthcare, and nutrition [15]. “Population density
[9–11], high degree of mobility of humans, and mass socialization, as well as cultural,
social, and tourism events [12–14] have been the basic reasons for COVID-19.” In this
description, in a nutshell, the main aim is to confer the effect of Work from Home in
rejoinder to the pestilence on education, income & wealth, and employment in India.
1.1 Education
From preschool to tertiary education, the education system has been affected, resultantly
worldwide policies have been introduced to target the complete shutdown of educational
institutions. Consequently, UNESCO estimated that this shutdown procedure of educa-
tional facilities has affected almost 900 million learners. At the same time as the objective
of these shutdowns is to prevent the spread of the virus and obviate carriage to defense-
less individuals in the institutions, these shutdowns have had ubiquitous socioeconomic
implications.
In the absence of a proper support system office, work, and household work, as
well as home time and school time, were inseparable during the lockdown and the
playtime for children became zero [16]. “Every house became a school and each parent
a teacher, during lockdown when schools and colleges were closed across India. There
was no boundary between the playtime and my time for millions of children in the
country. Further, it was realized the paucity of a structured learning environment at
home with having a worse impact on overall learning and consequently affected the
overall education outcome [16]”, education and SES are depicted in Fig. 1.
“As almost 70% of the 1.4 million schools and 51,000 colleges with nearly 300
million children are run by government bodies in India, the rural schools and the parents
now face a bleak education system and emptiness even as government’s advisories ask
schools to go online, and the government is looking at ways in which course can be
designed so students do not suffer.” The impact of a long-term school shutdown is yet
to be seen.
1.2 Employment
Many IT sector companies prefer WFH at a wide scale to enhance workplace flexibility
[17] and to reduce the worst impact on Society. The sudden importance and growth of
WFH have increased investigation of the WFH phenomenon, especially intending to
identify the number of jobs that can be done casually [18–22]. In general, the literature
overlooks the possible effects of WFH along with the unequal distribution of wages and
income. The causes of inequalities are multiple and distinct and have been growing in
eminence in policymakers, employment, and SES are depicted in Fig. 2.

102 P. Ojha et al.
According to Pouliakas and Branka (2020) and Fana et al. (2020), “the most defense-
less groups, such as women, non-natives, those with non-standard contracts (self-
employed and temporary workers), the lower educated, those employed in micro-sized
workplaces, and low-wage workers has been impacted by the COVID-19 pandemic.”
Consequently, Palomino et al. (2020) in their findings find that the crisis has increased
the levels of inequality and poverty [23]. Beland et al. (2020) examined “the short-term
consequences of COVID-19 on employment and wages as in his findings suggested that
the unemployment rate has been increased due to COVID-19; Working Hours and labor
force participation has decreased and had no significant impacts on wages [23].” Also,
this crisis has increased labor market inequalities. “According to the World Economic
Forum, the current pestilence compelled migrants to be trapped abroad and compromise
to the unfavorable circumstances, by taking up low-wage jobs, living in poor working
conditions, restricting spending, and thus, risk exposure to infections like the coronavirus
[24].”
1.3 Income and Wealth
Under our best observation, this study first shows how an increase in WFH would have an
impact on changes in income and wealth, as shown in Fig. 3. The lower socio-economic
stratum (SES) has been greatly affected by the economic downturn during the current
pandemic [15]. “The three main areas that have an economic impact of covid-19 are
given below:
•Elevation in poverty i.e., approaching more people below the poverty line [25]
•Aggravation of socio-economic disparities [26,27], and
•Conciliation in health-related precautions (use of masks, social distancing, looking
for medical guidance in case of cough and fever, etc.).”
PWFH OEO
Education SES
Fig. 1. Education and SES
In the current situation inequalities of income & wealth shocked younger households
and middle-aged households respectively. One of the disruptions which are caused by
this pandemic has had a major bang on the remittance flows used by migrant Indian

Impact of Work from Home During Covid-19 103
CWFH SE
EBS Employment
SES
Fig. 2. Employment and SES
workers; works as one of the ways of poverty diminution, economic development, and
boosting GDP. In India, remittances are anticipated to go down by about 23% in 2020;
with a remarkable gap to a growth of 5.5% in 2019 [28]. WFH system has emerged with
Covid-19, under which the people were suggested to work, study, and worship from
home.
TWFH BOBR
WG Income
SES
Fig. 3. Income and SES
Educationalists were also invited to adopt work from home system using technology,
as per the orders of The Ministry of Education. WFH for teachers has a few advantages
and disadvantages as well, for the performance of teachers. Also Work from home can

104 P. Ojha et al.
be carried out successfully if both the Educationalists and the educational institution go
through it dutifully [29]. Talking about certain disadvantages of WFH is that teachers
may not have any motivation to work due to a few constraints, like salary cuts, firing, etc.,
which reduce their income, consequently an aberration of enthusiasm and motivation.
Although WFH is considered the most effective way of performing activities, it helps to
minimize pestilence crisis and helps to run economic activities to earn Income.
Education
Employment
Income
SES
Fig. 4. SES model
In India comorbidity of this Pandemic has a great impact on Socioeconomic status
(SES), especially during lockdown and post-lockdown. The above model shown in Fig. 4
of SES represents the relationship between three variables education, employment, and
Income & wealth that are analyzed based on the independent variables, given below:
i. preference for work from home (PWFH),
ii. comfortable with work from home (CWFH),
iii. Time to work from home (TWFH)
and dependent variables are also given below:
i. online classes effective than offline (OEO),
ii. Self-employed (SE),
iii. economically beneficial for society (EBS)
iv. boost in online business revenue (BOBR) and
v. larger wealth gap (WG).
One of the most notable to this model is the socio-economic status of India is framed
by OEO, SE, EBS, BOBR, and the WG. During covid-19, the online classes were more

Impact of Work from Home During Covid-19 105
effective than Offline classes as children realized as per their safety basis with this
the online business revenue has increased as various activities have only one way to be
performed i.e., Online. People enjoyed lockdown with the help of online games and other
entertainment options, hence we can say lockdown enhanced the use of online platforms.
Due to this pandemic, people realized to have technical knowledge that again encouraged
cognitive behavior. Revenue from online businesses encouraged online employment
in the form of self-employment which could decrease unemployment. WFH also has
optimistic brim over effects on workers as it is beneficial to them for increased income
and reduced infection risks [18].
“As in the US economy [23], Beland et al. (2020) examined that covid-19 leads to an
increase in the unemployment rate, working hours, as well as the participation of the labor
force, has decreased; India faced the same issues due to which income and employment
level went down.” This happening allows other problems, like a larger wealth gap with
increased income inequalities and poverty to have emerged. Further, this increased the
scope for self-employment during the post-lockdown period under a good preview of the
SES (Socio-economic status) of India. For the growth of any economy like India, SE &
BOBR play a vital role to design a dignified SES. With this reference model of SES,
we examined the performance of the OEO, SE, BOBR & wealth gap in the landscape
of education, employment, and income & wealth to encourage the growth of SES in
India. Pandemic is responsible for shutting down certain employment opportunities,
decreased income sources, and more impact on education, but on the contrary, we found
certain development in these fields. Likewise, innovations are positively related to worse
conditions, as it is said in a worse situation when we have no more options, the human
mind conquers new ideas, and it leads to innovations. With these arguments, we analyzed
that SES is the outcome of alterations we have in OEO, SE, EBS, BOBR, and WG as
these were enhanced during this crisis.
2 Theoretical Background and Hypothesis Development
In this paper, we took 257 respondents from schools, Universities, professionals, industry
persons, and academicians from corner to corner via social media platform (WhatsApp)
in India to understand the effect of WFH on education, employment, and Income in India.
The data is limited to a few states like Uttar Pradesh, New Delhi, Uttarakhand, Maha-
rashtra, Gujarat, Punjab, Assam, Bihar, and West Bengal. Drawn from our arguments
and past research we developed three hypotheses. Under the surveillance of covid-19;
the study was conducted based on primary data collection (n =257) in online mode.
We selected University students, teachers, and other participants on a convenience sam-
pling basis to ensure feasibility. Quite a lot of advantages and disadvantages to the WFH
program have been observed by different researchers, as WFH activity is more flexible
than the physical activities to complete the work [29]. In education as well as in other
professions like IT sectors the stress level has been decreased with a distancing from
traffic jams and also have more free time for family. This gives a boost for the employees
to strengthen their ability.
Various research has confirmed that WFH is beneficial for the health of the country
socially and economically [30], hence we thought to go for an analysis of the Impact of

106 P. Ojha et al.
WFH on the SES of any nation, like India. For this, we tried to get information related
to the three elements of SES (education, health, and income) that define the health of
any nation.
We had a set of questions through an online survey anonymously, using the non-
probability snowball sampling technique that has been framed first on demographics
like age and gender; then the questions were divided into three parts
i. Education
ii. Employment
iii. Income & wealth.
For the first part of the questionnaire, we asked teachers & professors, do they feel
comfortable with online classes? and do these online classes are more effective than
offline? Further, the questionnaire consists of two questions that have been asked to
private employees (teachers & professors, low and middle-class workers, and Industry
persons) do they think work from home is economically beneficial for society? and Do
self-employment is the outcome of ‘work from home’ during this pandemic? Finally, we
asked three questions to them including estate dealers and purchasers; do they think it
is appropriate or suitable for the health of any nation? Do online movements facilitate a
boost in revenue from online businesses? do Wealth gaps (like income equality) become
larger during this Pandemic?
H1: Effect of WFH on Education regarding the independent variable PWFH and
dependent variable OEO.
H2: Impact of WFH on Employment regarding the Independent variable CWFH and
dependent variables SE and EBS.
H3: Impact of WFH on Income& wealth regarding the independent variable TWFH and
dependent variables BOBR and WG.
H4: SES depends on PWFH, CWFH, and TWFH with the special reference to education,
employment, and income & wealth.
The collected information was then analyzed by Simple Linear Regression analysis
in SPSS. We examined close relationships between different variables taken in the study.
Based on Descriptive Statistics, we found the Range =1, mean (n =257) =1.44, S.D. =
0.499 of all respondents.
A. Study 1: We took 100 students and teachers out of 257 respondents and found that
online study is more effective than offline as it reduces infection risks and enhanced
the technical knowledge of both. Further, the results i.e., P <0.002, R2 =0.131, and
F=10.266 stated that the overall regression model was significant. This has suggested
that students and teachers prefer online classes, consequently preferring WFH and so
contributing to the growth of SES, as shown in Tables 1,2, and 3.

Impact of Work from Home During Covid-19 107
Tabl e 1 . Model summary
Model R R square Adjusted R square Std.erroroftheestimate
1.360a.131 .118 .939
aPredictors: (Constant), PWFH.
R2 =0.131; taken as a set, the predictors i.e., dependent variables account for 13.1% of the
variance in the independent variable.
Tabl e 2 . ANOVAa(test using alpha =0.05)
Model Sum of squares Df Mean square F Sig.
1Regression 9.052 19.052 10.266 .002b
Residual 60.836 69 .882
Total 69.887 70
aDependent Variable: OEO.
bPredictors: (Constant), PWFH.
The overall regression model was significant, F =(9.052, 60.836) =10.266,
Tabl e 3 . Co-efficientsa(test each predictor at alpha =0.05)
Model Unstandardized
coefficients
Standardized coefficients t Sig.
B Std. error Beta
1 (Constant) 2.861 .274 10.457 .000
PWFH .376 .118 .360 3.204 .002
aDependent Variable: OEO.
B. Study 2: This study deals with the second hypothesis, where we found that EBS is
insignificant at P <0.221, R2 =0.023, but SE is significant with P <0.001, R2 =
0.157. Examining this we can state that self-employment has been encouraged during
Covid-19, on the contrary, WFH is not economically beneficial for society because of a
dearth of motivation, and competition and has hampered Industrial work (fieldwork), as
shown in Tables 4,5, and 6.
During this pestilence, self-employment has been encouraged due to less employ-
ment in the economy and cutting of salaries, which discouraged employees to remain in
the job. Although the business also had many constraints during this period, still people
were ready to engage themselves in business activities.
C. Study 3: An extrapolation of the below preliminary findings suggests that the first
variable in TWFH is ‘boost in online business revenue’ has no significant effect on SES.
From the results, we found P <0.725, R2 =0.002 which shows only 0.2% of the variance

108 P. Ojha et al.
Tabl e 4 . Model summary
Model R R square Adjusted R square Std.erroroftheestimate
1.147a.023 .008 1.071
aPredictors: (Constant), CWFH.
R2=.023; taken as a set, the predictors i.e., dependent variables account for 2.3% of the variance
in the independent variable.
Tabl e 5 . ANOVAa(test using alpha =0.05)
Model Sum of squares Df Mean square F Sig.
1Regression 1.755 11.755 1.528 .221b
Residual 79.203 69 1.148
Total 80.958 70
aDependent Variable: EBS.
bPredictors: (Constant), CWFH.
The overall regression model was significant, F =(1.755, 79.203) =1.528.
Tabl e 6 . Coefficientsa(test each predictor at alpha =0.05)
Model Unstandardized
coefficients
Standardized coefficients t Sig.
B Std. error Beta
1 (Constant) 2.090 .277 7.548 .000
CWFH .132 .107 .147 1.236 .221
aDependent Variable: EBS.
in the independent variable. Although during lockdown people at home preferred to play
online and also it has been observed predilection for online entertainment, as shown in
Tables 7,8,And9.
Tabl e 7 . Model summary
Model R R square Adjusted R square Std.erroroftheestimate
1.396a.157 .145 1.071
a.Predictors: (Constant), CWFH.
R2=0.157; taken as a set, the predictors i.e., dependent variables account for 15.7% of the
variance in the independent variable

Impact of Work from Home During Covid-19 109
Tabl e 8 . ANOVAa(test using alpha =0.05)
Model Sum of squares Df Mean square F Sig.
1Regression 14.740 114.740 12.842 .001b
Residual 79.203 69 1.148
Total 93.944 70
aDependent Variable: SE.
bPredictors: (Constant), CWFH.
Tabl e 9 . Coefficients (test each predictor at alpha =0.05)
Model Unstandardized
coefficients
Standardized coefficients t Sig.
B Std. error Beta
1(Constant) 1.090 .277 3.937 .000
CWFH .382 .107 .396 3.584 .001
aDependent Variable: SE.
The overall regression model was significant, F =(14.740, 79.203) =12.842.
D. Study 4: In this study, we examined the relationship between TWFH and WG to know
whether these variables are interconnected or not. Although we know that there is a very
close relationship but during the pandemic, income decreased at a remarkable rate and for
this reason, our analysis showed insignificant results and a low percentage of variance.
People need more time for WFH and the income to be increased; it is predicted that
WFH is preferred by It companies and others forever, in that case, Income will increase
and SES as well. R2 =0.005; taken as a set, the predictors i.e., dependent variables
account for0.5% of the variance in the independent variable, shown in Tables 10,11,12,
13,14, and 15.
Table 10. Model summary
Model R R square Adjusted R square Std.erroroftheestimate
1.043a.002 −.013 .855
aPredictors: (Constant), TWFH.
R2=0.002; taken as a set, the predictors i.e., dependent variables account for 0.2% of the variance
in the independent variable
The above analysis revealed that WG and BOBR have insignificant relations, but
both have a positive relationship with SES. Shreds of evidence from Tables 4and 5
explain the reason why WFH was one of the instruments in reducing infection rates
during the early days of the pestilence.

110 P. Ojha et al.
Table 11. ANOVAa(test using alpha =0.05)
Model Sum of squares Df Mean square F Sig.
1Regression .091 1.091 .125 .725b
Residual 50.387 69 .730
Total 50.479 70
aDependent Variable: BOBR.
bPredictors: (Constant), TWFH.
The overall regression model was significant, F =(0.091, 50.387) =12.842.
Table 12. Coefficientsa(test each predictor at alpha =0.05)
Model Unstandardized
coefficients
Standardized coefficients t Sig.
B Std. error Beta
1 (Constant) 1.541 .281 5.478 .000
TWFH .034 .097 .043 .354 .725
aDependent Variable: BOBR.
Table 13. Model summary
Model R R square Adjusted R square Std.erroroftheestimate
1.071a.007 −.010 1.236
aPredictors: (Constant), TWFH.
Table 14. ANOVAa(test using alpha =0.05)
Model Sum of squares Df Mean square FSig
1Regression .515 1.515 .337 .563b
Residual 105.401 69 1.528
Total 105.915 70
aDependent Variable: WG.
bPredictors: (Constant), TWFH.
The overall regression model was significant, F =(0.515, 105.401) =0 .337.
3 Conclusion
During the period of pestilence, we all are moving with a threat of being caught in
this trap of pandemic and don’t have any clues on how to get rid of the situation; we
are worried for our family and obviously for us too, knowing the adverse impact of

Impact of Work from Home During Covid-19 111
Table 15. Coefficientsa(test each predictor at alpha =0.05)
Model Unstandardized
coefficients
Standardized coefficients t Sig.
B Std. error Beta
1(Constant) 2.047 .407 5.032 .000
TWFH .081 .140 .070 .581 .563
aDependent Variable: WG.
covid-19. For the time being, we have vaccination now, but every higher authority has
question marks in their minds about whether they can solve this issue at that level of
desire of the public. Many efforts have been done to fight with covid-19, but not got
the final solution. In between that, every nation tried to overcome this issue at its best
levels. India also revealed the best part of its socio-economic aspects by balancing the
situation by applying WFH which is the utmost during the pandemic. This paper argued
that work from home is very much effective as it saves lives and the economy as well.
All else equal, the education, employment and income level of the economy have a
worse impact because of this pandemic and WFH allows reducing infection risk while
maintaining both economic and social activities. In this paper we took these (education,
employment, and income)three parts of SES as indicators and compared them with a
preference for work from home (PWFH), comfortable with work from home (CWFH), &
Time to work from home (TWFH) as independent variables; and dependent variables
i) online classes effective than offline (OEO), ii) Self-employed (SE), iii) economically
beneficial for society (EBS) iv) boost in online business revenue (BOBR) and v) larger
wealth gap(WG); to examine the relationships. The results were shocking for different
dependent variables, we found the significant relations of all to SES except one variable
i.e., WG which gave insignificant results during the first phase of covid-19. We examined
that WFH benefited the socio-economic part of the nation with few negative impacts that
imply WFH should be encouraged as long as noteworthy virus risk remains.
References
1. Sonia, S.: Pandemic: Tracking Contagions, From Cholera to Ebola and Beyond. Sarah
Crichton Books, New York (2016). ISBN 978-0-374-12288-1
2. Saini, S.: COVID-19 may double poverty in India [Internet] Financial Express 2020.
https://www.financialexpress.com/opinion/covid-19-may-double-poverty-in-india/194
3736/. Accessed 22 May 2020
3. Purwanto, A., et al.: Impact of work from home (WFH) on Indonesian teachers performance
during the Covid-19 pandemic: an exploratory study. Int. J. Adv. Sci. Technol. 29(5), 6235–
6244 (2020)
4. Bick, A., Blandin, A., Mertens, K.: Work from home after the Covid-19 outbreak. CEPR
Discussion Paper No. DP15000
5. Alon, T., Doepke, M., Rumsey, J.-O., Tertilt, M.: The impact of COVID-19 on gender equality.
In: NBERvWorking Papers 26947. Inc, National Bureau of Economic Research (2020)

112 P. Ojha et al.
6. Anser, M.K., Yousaf, Z., Khan, M.A., Nassani, A.A., Alotaibi, S.M., Qazi Abro, M.M., et al.:
Does communicable diseases (including COVID-19) may increase global poverty risk? A
cloud on the horizon. Environ. Res. 15(187), 109668 (2020)
7. Atkeson, A.: What will be the Economic Impact of COVID-19 in the US? Rough Estimates of
Disease Scenarios. National Bureau of Economic Research, Cambridge (2020). http://www.
nber.org/Papers/w26867. Accessed 21 June 2020
8. Baker, S., Bloom, N., Davis, S., Kost, K., Sammon, M., Viratyosin, T.: The Unprecedented
Stock Market Impact of COVID-19. https://www.nber.org/papers/w26945. Accessed May 21
2020
9. Bartik, A., Bertrand, M., Cullen, Z., Glaeser, E., Luca, M., Stanton, C.: How Are Small Busi-
nesses Adjusting to COVID19? Early Evidence from a Survey. National Bureau of Economic
Research, Cambridge (2020). https://www.nber.org/papers/w26989. Accessed 21 June 2020
10. Béland, L.-P., Brodeur, A., Wright, T.: The short-term economic consequences of COVID-
19: exposure to disease, remote work and government response. IZA Discussion Paper Series
(13159) (2020)
11. Di Gennaro, F., et al.: Coronavirus diseases (COVID-19) current status and future perspectives:
a narrative review. Int. J. Environ. Res. Public Health 17, 2690 (2020).
12. Dingel, J., Neiman, B.: How many jobs can be done at home? National Bureau of Economic
Research No. 26948 (2020)
13. Guerrieri, V., Lorenzoni, G., Straub, L., Werning, I.: Macroeconomic Implications of COVID-
19: Can Negative Supply Shocks Cause Demand Shortages? National Bureau of Economic
Research, Cambridge (2020). http://www.nber.org/papers/w26918.pdf. Accessed 21 June
2020
14. Guermond, V., Datta, K.: How coronavirus could hit the billions migrant workers send home
[Internet] World Economic Forum 2020. https://www.weforum.org/agenda/2020/04/how-cor
onavirus-could-hitthe-Billions-migrant-workers-send-home/. Accessed 23 Apr 2020
15. Mccloskey, B., et al.: Mass gathering events and reducing further global spread of COVID-19:
a political and publicHealth dilemma. Lancet 395, 1096–1099 (2020)
16. Bonacini, L., Gallo, G., Scicchitano, S.: Working from home and income inequality: risks of
a ‘new normal’ with COVID-19. J. Popul. Econ. 34, 303–360 (2021)
17. Mahendradev, S.: Addressing COVID-19 impacts on agriculture, food security,
and livelihoods in India | IFPRI: international food policy research institute.
IFPRI. https://www.ifpri.org/Blog/addressing-covid-19-impacts-agriculture-food-security-
and-livelihoods-India. Accessed 22 May 2020
18. Alipour, J.V., Fadinger, H., Schymik, J.: My home is my castle – the benefits of working from
home during a pandemic crisis. J. Public Econ. 196, 104373 (2021)
19. Kang, D., Choi, H., Kim, J.H., Choi, J.: Spatial epidemic dynamics of the COVID-19 outbreak
in China. Int. J. Infect. Dis. 94, 96–102 (2020)
20. Koren, M., Peto, R.: Business disruptions from social distancing. In: Covid Economics (2),
13–31. Press, CEPRIZA Discussion Paper No. 1328133 Pages Posted, 23 May 2020 Kon-
stantinos Pouliakas Cedefop, University of Aberdeen - Business School, IZA Institute of
Labor Economics Jiri Branka (2020)
21. Leibovici, F., Santacrue, A.M., Famiglietti, M.: Social distancing and contact-intensive
occupations. St. Louis Federal Reserve Bank - On the Economy Blog, March (2020)
22. Lemay, M.C.: Global Pandemic Threats: A Reference Handbook. ABC-CLIO, Santa Barbara
(2020). ISBN 978-1-4408-4282-5
23. Honigsbaum, M.: The Pandemic Century: One Hundred Years of Panic, Hysteria, and Hubris.
W. W. Norton & Company, New York (2019). ISBN 978-0393254754
24. Ito, H., Hanaoka, S., Kawasaki, T.: The cruise industry and the COVID-19 outbreak. Transp.
Res. Interdiscip. Perspect. 5, 100136 (2020)

Impact of Work from Home During Covid-19 113
25. Mongey, S., Pilossoph, L., Weinberg. A.: Which workers bear the burden of social distancing
policies? NBER Working Paper No. 27085 (2020)
26. Fana, M., Torrejon Prez, S., Fernandez-Macias, E.: Employment impact of Covid-19: from
short term effects to long terms prospects. J. Ind. Bus. Econ. 47, 391–410 (2020)
27. Pouliakas, K., Branka, J.: EU Jobs at highest risk of Covid-19 social distancing: will the
pandemic exacerbate labour market divide? IZA discussion paper no. 13281. https://ssrn.
com/abstract=3608530
28. Praharaj, S., Vaidya, H.: The Urban Dimension of COVID-19 in India: COVID Outbreak
and Lessons for Future Cities. https://www.researchgate.net/publication/341616744_The_
urban_dimension_of_COVID19_in_India_COVID_Outbreak_and_Lessons_for_Future_Cit
ies?Channel=doi&Linkid=5ecb837492851c11a8880043&showfulltext=true. Accessed 21
May 2020
29. Prashant, K.N., Khanna, P.: Every house a school, every parent a teacher as Covid-19 impacts
education of 300mn students. https://www.livemint.com/news/india/every-house-a-school-
every-parent-a-teacher-as-covid-19-impacts-education-11585140662556.html
30. Adams-Prassl, A., Boneva, T., Golin, M., Rauh, C.: Inequality in the impact of the coronavirus
shock: evidence from real-time surveys. IZA Discussion Paper No. 13183 (2020)

Author Index
Ali, Arif 3
Arjun, K. P. 89
Awasthi, Monisha 22
Awasthi, Prakhar 22
Basak, Arindam 73
Chakravarty, Debashish 73
Chawla, Priyanka 63
Choudhury, Akash Sur 73
Gairola, Ajay Krishan 51
Garg, Sharvan Kumar 3
Goel, Ankur 22
Goyal, Tushar 16
Halder, Tamesh 73
Husain, Arshad 16
Jangir, Yuvraj 16
Kandari, Sumit 16
Khanduja, Manisha 22
Kumar, Anuj 22
Kumar, Vidit 51
Maurya, Sudhanshu 100
Mishra, Anupama 39
Nagaraju, M. 63
Ojha, Manish Kumar 100
Ojha, Poonam 100
Pandey, Akhilesh 3
Rawat, Deepesh 39
Rawat, Tushar 89
Saini, Shivani 3
Sajwan, Vijaylakshmi 22
Singh, Pankaj Pratap 3
Singh, Rohan 89
Sreenarayanan, N. M. 89
Tiwari, Rajeev 63