Machine Learning In Information And Communication Technology Proceedings Of Icict 2021 Smit Hiren Kumar Deva Sarma

Machine Learning In Information And
Communication Technology Proceedings Of Icict
2021 Smit Hiren Kumar Deva Sarma download
https://ebookbell.com/product/machine-learning-in-information-
and-communication-technology-proceedings-of-icict-2021-smit-
hiren-kumar-deva-sarma-47288552
Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.
Deployable Machine Learning For Security Defense Second International
Workshop Mlhat 2021 Virtual Event August 15 2021 Proceedings
Communications In Computer And Information Science 1st Ed 2021 Gang
Wang Editor
https://ebookbell.com/product/deployable-machine-learning-for-
security-defense-second-international-workshop-mlhat-2021-virtual-
event-august-15-2021-proceedings-communications-in-computer-and-
information-science-1st-ed-2021-gang-wang-editor-38451846
Android Malware Detection Using Machine Learning Datadriven
Fingerprinting And Threat Intelligence 86 Advances In Information
Security 86 1st Ed 2021 Elmouatez Billah Karbab
https://ebookbell.com/product/android-malware-detection-using-machine-
learning-datadriven-fingerprinting-and-threat-
intelligence-86-advances-in-information-security-86-1st-
ed-2021-elmouatez-billah-karbab-33377426
Android Malware Detection Using Machine Learning Datadriven
Fingerprinting And Threat Intelligence 86 Advances In Information
Security 86 Elmouatez Billah Karbab Mourad Debbabi Abdelouahid Derhab
Djedjiga Mouheb
https://ebookbell.com/product/android-malware-detection-using-machine-
learning-datadriven-fingerprinting-and-threat-
intelligence-86-advances-in-information-security-86-elmouatez-billah-
karbab-mourad-debbabi-abdelouahid-derhab-djedjiga-mouheb-33377428
Artificial Intelligence And Machine Learning 33rd Benelux Conference
On Artificial Intelligence Bnaicbenelearn 2021 Eschsuralzette
Luxembourg In Computer And Information Science 1530 1st Edition Luis A
Leiva
https://ebookbell.com/product/artificial-intelligence-and-machine-
learning-33rd-benelux-conference-on-artificial-intelligence-
bnaicbenelearn-2021-eschsuralzette-luxembourg-in-computer-and-
information-science-1530-1st-edition-luis-a-leiva-37518338

Artificial Intelligence And Machine Learning 32nd Benelux Conference
Bnaicbenelearn 2020 Leiden The Netherlands November 1920 2020 Revised
In Computer And Information Science 1398 1st Edition Mitra Baratchi
https://ebookbell.com/product/artificial-intelligence-and-machine-
learning-32nd-benelux-conference-bnaicbenelearn-2020-leiden-the-
netherlands-november-1920-2020-revised-in-computer-and-information-
science-1398-1st-edition-mitra-baratchi-34627412
Availability Reliability And Security In Information Systems Ifip Wg
84 89 Tc 5 International Crossdomain Conference Cdares 2016 And
Workshop On Privacy Aware Machine Learning For Health Data Science
Paml 2016 Salzburg Austria August 31 S Francesco Buccafurri
https://ebookbell.com/product/availability-reliability-and-security-
in-information-systems-ifip-wg-84-89-tc-5-international-crossdomain-
conference-cdares-2016-and-workshop-on-privacy-aware-machine-learning-
for-health-data-science-paml-2016-salzburg-austria-
august-31-s-francesco-buccafurri-5607748
Handbook Of Research On Machine And Deep Learning Applications For
Cyber Security Advances In Information Security Privacy And Ethics
Padmavathi Ganapathi
https://ebookbell.com/product/handbook-of-research-on-machine-and-
deep-learning-applications-for-cyber-security-advances-in-information-
security-privacy-and-ethics-padmavathi-ganapathi-11069536
Exercises In Applied Mathematics With A View Toward Information Theory
Machine Learning Wavelets And Statistical Physics Daniel Alpay
https://ebookbell.com/product/exercises-in-applied-mathematics-with-a-
view-toward-information-theory-machine-learning-wavelets-and-
statistical-physics-daniel-alpay-57081340
Applications Of Machine Learning And Artificial Intelligence In
Education Seda Khadimally
https://ebookbell.com/product/applications-of-machine-learning-and-
artificial-intelligence-in-education-seda-khadimally-48752580

Lecture Notes in Networks and Systems 498
Hiren Kumar Deva Sarma
Vincenzo Piuri
Arun Kumar Pujari Editors
Machine
Learning in
Information and
Communication
Technology
Proceedings of ICICT 2021, SMIT

Lecture Notes in Networks and Systems
Volume 498
Series Editor
Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences,
Warsaw, Poland
Advisory Editors
Fernando Gomide, Department of Computer Engineering and Automation—DCA,
School of Electrical and Computer Engineering—FEEC, University of
Campinas—UNICAMP, São Paulo, Brazil
Okyay Kaynak, Department of Electrical and Electronic Engineering,
Bogazici University, Istanbul, Turkey
Derong Liu, Department of Electrical and Computer Engineering, University of
Illinois at Chicago, Chicago, USA
Institute of Automation, Chinese Academy of Sciences, Beijing, China
Witold Pedrycz, Department of Electrical and Computer Engineering, University of
Alberta, Alberta, Canada
Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland
Marios M. Polycarpou, Department of Electrical and Computer Engineering,
KIOS Research Center for Intelligent Systems and Networks, University of Cyprus, Nicosia, Cyprus
Imre J. Rudas, Óbuda University, Budapest, Hungary
Jun Wang, Department of Computer Science, City University of Hong Kong,
Kowloon, Hong Kong

The series “Lecture Notes in Networks and Systems” publishes the latest
developments in Networks and Systems—quickly, informally and with high quality.
Original research reported in proceedings and post-proceedings represents the core
of LNNS.
Volumes published in LNNS embrace all aspects and subﬁelds of, as well as new
challenges in, Networks and Systems.
The series contains proceedings and edited volumes in systems and networks,
spanning the areas of Cyber-Physical Systems, Autonomous Systems, Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace Systems,
Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power Systems, Robotics, Social Systems, Economic Systems and other. Of particular value to both the contributors and the readership are the short publication timeframe and
the world-wide distribution and exposure which enable both a wide and rapid dissemination of research output.
The series covers the theory, applications, and perspectives on the state of the art
and future developments relevant to systems and networks, decision making, control, complex processes and related areas, as embedded in the ﬁelds of interdisciplinary
and applied sciences, engineering, computer science, physics, economics, social, and life sciences, as well as the paradigms and methodologies behind them.
Indexed by SCOPUS, INSPEC, WTI Frankfurt eG, zbMATH, SCImago.
All books published in the series are submitted for consideration in Web of Science.
For proposals from Asia please contact Aninda Bose ([email protected]).

Hiren Kumar Deva Sarma
Arun Kumar Pujari
Editors
Machine Learning
in Information
and Communication
Technology
Proceedings of ICICT 2021, SMIT

Editors
Hiren Kumar Deva Sarma
Department of Information Technology
Gauhati University
Jalukbari, Assam, India
Arun Kumar Pujari
Department of Computer Science and Engineering Mahindra University Hyderabad Hyderabad, India
Vincenzo Piuri
University of Milan Milan, Italy
ISSN 2367-3370 ISSN 2367-3389 (electronic)
Lecture Notes in Networks and Systems
ISBN 978-981-19-5089-6 ISBN 978-981-19-5090-2 (eBook)
https://doi.org/10.1007/978-981-19-5090-2
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2023
This work is subject to copyright. All rights are solely and exclusively licensed 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 Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface
Information and communication technologies have grown up signiﬁcantly and pene-
trated in almost all dimensions of human society. There has been tremendous growth
in all the areas of information technology including Information gathering, informa-
tion processing, information sharing, information distribution, information retrieval,
information security, and so on. Internet has revolutionized the human society and
opened up many new dimensions that include information sharing models, elec-
tronic commerce, and also new economic models altogether. Although artiﬁcial
intelligence was conceptualized long time back, at present time, there has been a
tendency to achieve wide meaningful practice of artiﬁcial intelligence in all major
domains of knowledge, even apart from information technology-based applications.
Machine learning (ML) algorithms are playing important role in this aspect. Overall,
health care is such a domain of knowledge at present, in which ML algorithms are
being widely applied for better processing of healthcare data in order to make mean-
ingful decisions. Penetration of information technology in medical and healthcare
domain is worth noticing, which is expected to help health professionals signiﬁcantly
in making wise and better decisions. Apart from healthcare data-related studies,
application of ML algorithms is expanding and inﬂuencing all other domains of
information and communication technologies, in general. Communication technolo-
gies are getting enhanced signiﬁcantly and modern communication paradigms are
emerging and evolving. In the area of communication, optimization in terms of
energy expenditure, overall cost of communication and even life span of communi-
cation systems has always been an issue of great signiﬁcance. In recent time, it has
been witnessed that there is a growing interest in applying ML algorithms in various
parts of communication systems in search of better solutions.
Therefore, in these book, effort has been made to see the three major contemporary
areas of knowledge, namely information technology, communication technology and machine learning, in a connected manner. The aim of the book is to share the latest
research happenings in applying ML algorithms in different information technology- based applications as well as in communication technologies.
v

vi
There are 31 research articles in this book. These are the research papers presented
during the ﬁrst International Conference on Information and Communication Tech-
nology (ICICT) 2021 held at Sikkim Manipal Institute of Technology (SMIT) during
23–24 December 2021 organized by the Department of Information Technology. The
articles are broadly categorized into the following: healthcare informatics, recom-
mendation systems, communication networks, assistive technology, social networks,
image and video processing, cybersecurity, and miscellaneous applications.
We are thankful to all the authors of ICICT 2021. We sincerely thank SMIT for
extending all the support in organizing ICICT 2021. We are highly grateful to Prof. Janusz Kacprzyk for his guidance and encouragement throughout. We sincerely thank
the editorial team of Springer Nature for their great support that we have received. We sincerely thank Springer Nature for being our publication partner and for all the help and great support that have been extended to us during the making of this book.
Jalukbari, India
Milan, Italy
Hyderabad, India
Hiren Kumar Deva Sarma
Vincenjo Piuri
Arun Kumar Pujari

Contents
Healthcare Informatics
Earth Mover’s Distance-Based Automated Disease Tagging
of Indian ECGs...................................................3
Burhan Basha, Dhruva Nandi, Karuna Nidhi Kaur,
Priyadarshini Arambam, Shikhar Gupta, Mehak Segan, Priya Ranjan,
Upendra Kaul, and Rajiv Janardhanan
Cardial Disease Prediction in Multi-variant Systems Using
MT-MrSBC Model................................................21
Pandiyan Nandakumar and Subhashini Narayan
SVM-based Pre- and Post-treatment Cancer
Segmentation from Lung and Abdominal CT Images
via Neighborhood-Inﬂuenced Features..............................35
Tiyasa Chakraborty, Ashok Kumar Bhadra, and Debashis Nandi
Blood Cancer Detection with Microscopic Images Using Machine
Learning.........................................................45
Christo Ananth, P. Tamilselvi, S. Agnes Joshy, and T. Ananth Kumar
A Survey on Machine Learning-Based Approaches for Leukaemia
Detection.........................................................55
Leena I. Sakri and Rajeshwari V. Patil
Periocular Region Recognition—A Brief Survey......................63
R. Sheela and R. Suchithra
A Bibliometric Analysis on the Relationship Between Emotional
Intelligence, Self-Management and Health Information Seeking.......77
Jennifer Gurung, Vivek Pandey, Samrat Kumar Mukherjee,
Saibal Kumar Saha, Ankit Singh, and Ajeya Jha
vii

viii
Preferences in the Detailing Process Among Young and Senior
Physicians........................................................89
Saibal Kumar Saha, Ankita Sarangi, Sonia Munjal, Piyanka Dhar,
and Ajeya Jha
Non-cognitive Differences on Social Media Branded Drugs
Promotion: Study of Indian Patients and Physicians..................97
Samrat Kumar Mukherjee, Jitendra Kumar, Jaya Rani Pandey,
Vivek Chhetri, and Ajeya Jha
AUTCD-Net: An Automated Framework for Efﬁcient Covid-19
Diagnosis on Computed Tomography Scans..........................109
Palash Ghosal, Amish Kumar, Soumya Snigdha Kundu,
Utkarsh Prakash Srivastava, Ashis Datta, and Hiren Kumar Deva Sarma
Computer-Aided Detection of Brain Midline Using CT Images.........117
Palash Ghosal, Amish Kumar, Ashis Datta, Hiren Kumar Deva Sarma,
and Debashis Nandi
Deep Neural Network-Based Classiﬁcation of ASD
and Neurotypical Subjects Using Functional Connectivity
Features Derived from Resting-State fMRI Data.....................125
Nirmal Rai, P. C. Pradhan, Hemanta Saikia, O. P. Singh, and Rinkila Bhutia
Application of Deep Learning in Healthcare..........................131
Aryan Shahi, Chandralika Chakraborty, Shubhodeep Ghosh, and Ankit Anand
Recommendation Systems
A Popularity-Based Recommendation System Using Machine
Learning.........................................................143
Pranati Rakshit, Sougata Saha, Arindam Chatterjee, Subhayan Mistri,
Swagata Das, and Gunjan Dhar
Machine Learning-Based Movies and Shows Recommendation
System...........................................................151
Sanmoy Dev Purkayastha, Suraj Kumar, Ashish Saha, and Saumya Das
Communication Networks
A Survey on Application of LSTM as a Deep Learning Approach
in Trafﬁc Classiﬁcation for SDN....................................161
Prerna Rai and Hiren Kumar Deva Sarma

Contents
Assistive Technology
A Novel Low-cost Visual Aid System for the Completely Blind
People............................................................177
Christo Ananth, M. Kameswari, R. Srinivasan, S. Surya,
and T. Ananth Kumar
Social Networks
Twitter Sentiment Analysis Using Machine Learning Techniques:
A Case Study Based on Farmers Protest.............................187
C. Sahithi, Y. Sreeja, S. Akhil, K. Taruni, and C. C. Sobin
Sentiment Analysis to Find Sentence Polarity on Tweet Data...........197
Pranati Rakshit, Sumit Gupta, and Tarpan Das
Impact of Emotional Support and Medical Adherence on Social
Media Branded Drug Promotion....................................203
Samrat Kumar Mukherjee, Jitendra Kumar, Ajeya Jha,
Jaya Rani Pandey, and Saibal Kumar Saha
Image and Video Processing
A Survey on Video Description and Summarization Using Deep
Learning-Based Methods...........................................213
Pranati Rakshit, Anuj Kumar, and Amlan Chakraborty
Aerial Image Classiﬁcation Using Convolution Neural Network........225
Praveen Kumar Pradhan and Udayan Baruah
Image Retrieval Using Neural Networks for Word Image
Spotting—A Review...............................................243
Naiwrita Borah and Udayan Baruah
Cybersecurity
Priority-Based Mitigation in Education Sector using Machine
Learning.........................................................271
Sonal Shukla and Anand Sharma
Miscellaneous Applications
Prediction and Analysis of Air Quality Index Using Machine
Learning Algorithms..............................................281
Avishek Choudhuri, R. Sujatha, Chhazed Shreyans Nitin, Jyotir Moy Chatterjee, and R. N. Thakur
A Short Overview on Various Bio-Inspired Algorithms................295
K. Boopalan, C. Shanmuganathan, K. Lokeshwaran, and T. Balaji

x
Gesture-Based Drawing Application: A Survey.......................303
Leena I. Sakri, Vijeta Kerur, Gautam Shet, Sohail Mokashi,
and Abhishek Patki
Simulation and Modeling of Electrical Load Data Using Machine
Learning.........................................................311
Manish Kumar and Nitai Pal
Analysis of Single Missing Gate Faults in Quantum Circuit............317
Shubhrojit Paul, Mousum Handique, and Hiren Kumar Deva Sarma
Experimental Validation of Frequency Scaling Over Indian Hilly
Region...........................................................327
Badichapta Deka Baro and Swastika Chakraborty
School Uniform Identiﬁcation Using Deep Learning Based
Approach.........................................................333
Ashis Datta, Sanju Kumar Giri, Vibhuti Sharma, Anushka Das,
and Joyashri Basak
Author Index......................................................341

Editors and Contributors
About the Editors
Prof. Hiren Kumar Deva Sarma
Technology, Gauhati University, Assam, India. He received Bachelor of Engineering
in Mechanical Engineering from Assam Engineering College, Guwahati, Assam
(1998). He completed Master of Technology in Information Technology from Tezpur
University, Assam (2000). He received Doctor of Philosophy (in Computer Science
and Engineering) from Jadavpur University, West Bengal (2013). He has co-authored
three books, edited four book volumes, and published more than eighty research
papers in different International Journals and refereed International and National
Conferences of repute. He is the recipient of the Young Scientist Award from Inter-
national Union of Radio Science (URSI) in the XVIII General Assembly 2005, held
at New Delhi, India and has received IEEE Early Adopter Award in 2014. His current
research interests are networks, network security, robotics and big data analytics.
Prof. Vincenzo Piuri guished Scientist. He is known for his work in the ﬁeld of information processing, with speciﬁc focus on artiﬁcial intelligence, computational intelligence, signal/image
processing, biometrics, industrial applications, measurement systems, arithmetic units and fault-tolerant architectures. He received his M.S. and Ph.D. in Computer Engineering from Politecnico di Milano, Italy. He is Full Professor at the Univer-
sity of Milan, Italy (since 2000), where he was also Department Chair (2007–2012). He was Associate Professor at Politecnico di Milano, Italy (1992–2000), Visiting Professor at the University of Texas at Austin, USA (summers 1996–1999), and
visiting researcher at George Mason University, USA (summers 2012–2019). He founded a start-up company, Sensure srl, in the area of intelligent systems for indus- trial applications (leading it from 2007 until 2010) and was active in industrial
research projects with several companies.
xi

xii Editors and Contributors
Prof. Arun Kumar Pujari
and Information Sciences at the University of Hyderabad (UoH). He has been Vice-
Chancellor of the Central University of Rajasthan (2015–2020) and Vice-Chancellor
of Sambalpur University, Odisha (20082011). He joined the UoH as Reader in 1985
and became a professor in 1990. Prior to joining University of Hyderabad, he served
JNU, New Delhi. He completed his postgraduation from the Sambalpur University
in 1974 and got his Ph.D. from IIT-Kanpur in 1980. He has more than 25 years’
experience as Vice-Chancellor, Dean, Head of Department, and other administrative
positions. He has served several high-level bodies of UGC, DST, DRDO, ISRO and
AICTE. He has wide exposure in the national and international arena, and he has
been invited to University of Tokyo, Japan, University of Paris-Sud, France, Grifﬁth
University, Australia, Memphis University, USA and IIST, United Nations Univer-
sity, Macau on different visiting assignments. He has more than 100 publications to
his credit and his book, Data Mining Techniques, was a popular textbook adopted
by several universities in India and abroad.
Contributors
Akhil S.
Anand Ankit
Ananth Christo
University, Samarkand, Uzbekistan
Arambam Priyadarshini
Noida, India
Balaji T. Tamil Nadu, India
Baro Badichapta Deka nology, Majitar, East Sikkim, India
Baruah Udayan of Technology, Sikkim Manipal University, Majitar, East Sikkim, India; Department of Information Technology, Sikkim Manipal Institute of Technology,
Sikkim Manipal University, Rangpo, Sikkim, India
Basak Joyashri
Bengal, India
Basha Burhan School of Engineering and Applied Sciences, SRM University, Amaravati, Andhra
Pradesh, India
Bhadra Ashok Kumar

Editors and Contributors xiii
Bhutia Rinkila
Boopalan K.
Andhra Pradesh, India
Borah Naiwrita of Technology, Sikkim Manipal University, Rangpo, Sikkim, India
Chakraborty Amlan
sity of Calcutta, Kolkata, India
Chakraborty Chandralika Majhitar, India
Chakraborty Swastika nology, Majitar, East Sikkim, India
Chakraborty Tiyasa
Institute of Technology Durgapur, Durgapur, India
Chatterjee Arindam College of Engineering, Kalyani, India
Chatterjee Jyotir Moy
Chhetri Vivek
Technology, Sikkim Manipal University, Gangtok, India
Choudhuri Avishek
Das Anushka
Sikkim, India
Das Saumya Technology, Sikkim Manipal University, Sikkim, India
Das Swagata Engineering, Kalyani, India
Das Tarpan
Engineering, Kalyani, India
Datta Ashis Sikkim Manipal Institute of Technology, Majitar, Rangpo, East Sikkim, India
Dhar Gunjan Engineering, Kalyani, India
Dhar Piyanka
sity, Gangtok, Sikkim, India
Ghosal Palash

xiv Editors and Contributors
Ghosh Shubhodeep
India
Giri Sanju Kumar
Sikkim, India
Gupta Shikhar
Gupta Sumit
Engineering, Kalyani, India
Gurung Jennifer sity, Gangtok, India
Handique Mousum
Janardhanan Rajiv
Health Sciences, SRM Institute of Science & Technology, Chennai, India
Jha Ajeya Technology, Sikkim Manipal University, Gangtok, Sikkim, India
Joshy S. Agnes India
Kameswari M.
Kalasalingam Academy of Research and Education, Srivilliputhur, Tamilnadu, India
Kaul Upendra Medical Research Centre, New Delhi, India
Kaur Karuna Nidhi
ology, Amity Institute of Public Health, Noida, India
Kerur Vijeta of Engineering and Technology, Dharwad, Karnataka, India
Kumar Anuj Bengaluru, India
Kumar Jitendra Technology, Sikkim Manipal University, Gangtok, India
Kumar Manish Mines), Dhanbad, India; Department of Nuclear Science and Technology, Pandit Deendayal Energy Univer-
sity, Gandhinagar, India
Kumar Suraj
Technology, Sikkim Manipal University, Sikkim, India
Kumar T. Ananth
College of Engineering, Villupuram, India

Editors and Contributors xv
Kundu Soumya Snigdha
lathur, India
Lokeshwaran K.
Melvisharam, Tamil Nadu, India
Mistri Subhayan of Engineering, Kalyani, India
Mokashi Sohail College of Engineering and Technology, Dharwad, Karnataka, India
Mukherjee Samrat Kumar Institute of Technology, Sikkim Manipal University, Gangtok, India
Munjal Sonia Uttarakhand, India
Nandakumar Pandiyan
Vellore Institute of Technology, Vellore, Tamil Nadu, India
Nandi Debashis Institute of Technology Durgapur, Durgapur, India
Nandi Dhruva Sciences, SRM Institute of Science & Technology, Chennai, India
Narayan Subhashini
Institute of Technology, Vellore, Tamil Nadu, India
Nitin Chhazed Shreyans
Pal Nitai
Dhanbad, India
Pandey Jaya Rani of Technology, Sikkim Manipal University, Gangtok, India
Pandey Vivek Gangtok, India
Patil Rajeshwari V.
Karnataka, India
Patki Abhishek College of Engineering and Technology, Dharwad, Karnataka, India
Paul Shubhrojit
Pradhan P. C.
India
Pradhan Praveen Kumar Chisopani, South Sikkim, India

xvi Editors and Contributors
Purkayastha Sanmoy Dev
Manipal Institute of Technology, Sikkim Manipal University, Sikkim, India
Rai Nirmal
Rai Prerna
Rakshit Pranati
of Engineering, Kalyani, India
Ranjan Priya
Saha Ashish Technology, Sikkim Manipal University, Sikkim, India
Saha Saibal Kumar
of Technology, Sikkim Manipal University, Gangtok, Sikkim, India
Saha Sougata Engineering, Kalyani, India
Sahithi C.
Saikia Hemanta
India
Sakri Leena I. of Engineering and Technology, Dharwad, Karnataka, India
Sarangi Ankita
sity, Gangtok, Sikkim, India
Deva Sarma Hiren Kumar Rangpo, East Sikkim, Sikkim, India
Segan Mehak Amity Institute of Public Health, Noida, India
Shahi Aryan
Shanmuganathan C.
Chennai, India
Sharma Anand
Sharma Vibhuti Sikkim, India
Sheela R. (Deemed-to-be University), Bangalore, India
Shet Gautam of Engineering and Technology, Dharwad, Karnataka, India
Shukla Sonal

Editors and Contributors xvii
Singh Ankit
Singh O. P.
Sobin C. C.
Sreeja Y.
Srinivasan R.
and Technology, Kattankulathur, India
Srivastava Utkarsh Prakash Sikkim, India
Suchithra R.
(Deemed-to-be University), Bangalore, India
Sujatha R.
Surya S.
Academy of Research and Education, Krishnan Koil, Srivilliputtur, Tamilnadu, India
Tamilselvi P. (Deemed to Be) University, Bangalore, India
Taruni K.
Thakur R. N.

Healthcare Informatics

Earth Mover’s Distance-Based
Automated Disease Tagging of Indian
ECGs
Burhan Basha, Dhruva Nandi, Karuna Nidhi Kaur,
Priyadarshini Arambam, Shikhar Gupta, Mehak Segan, Priya Ranjan,
Upendra Kaul, and Rajiv Janardhanan
Abstract Ours is the era of cardiovascular disorders. In this work, a corpus of ECGs
collected in the state of Jammu and Kashmir is studied for automated perceptual
similarity-based disease tagging. Following on our earlier work, we have deployed
Earth Mover’s distance (EMD) as the similarity metric to generate automated disease
tags. Rationale for generating these tags is based on the similarity of test ECG with
B. Basha
Department of Electronics and Communication Engineering, School of Engineering and Applied
Sciences,
SRM University, Amaravati, Andhra Pradesh, India
e-mail: [email protected]
D. Nandi · R. Janardhanan
Department of Medical Research, Faculty of Medical and Health Sciences, SRM Institute of
Science
& Technology, Kattankulathur, Chennai, India
e-mail: [email protected]
R. Janardhanan
e-mail: [email protected]
K. N. Kaur · M. Segan
Laboratory of Disease Dynamics and Molecular Epidemiology, Amity Institute of Public Health,
Noida,
India
e-mail: [email protected]
M. Segan
e-mail: [email protected]
P. A r a m b a m · S. Gupta
Amity Institute of Public Health, Amity University, Noida, India
e-mail: [email protected]
S. Gupta
e-mail: [email protected]
P. R a n j a n (
B)
ECE, Bhubaneshwar Institute of Technology, Info Valley, Harapur, Odisha, India
e-mail: [email protected]
U. Kaul
Batra Heart Centre, Academics and Research, Batra Hospital and Medical Research Centre, New
Delhi,
India
e-mail: [email protected]
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023
H.
K. Deva Sarma et al. (eds.), Machine Learning in Information and Communication
Technology, Lecture Notes in Networks and Systems 498,
https://doi.org/10.1007/978-981-19-5090-2_1
3

4
healthy and unhealthy ECGs. If the test ECG resembles representative healthy ECG,
then it is tagged as healthy, and if it resembles representative unhealthy ECG, then
it is tagged as unhealthy. Future directions for increasing the accuracy of this work
are discussed. It’s integration with biomarkers in a multi-sensor data fusion-based
automated CVD tagging criteria is also explored.
Keywords ·Electrocardiogram ·Disease tagging ·
Cardiovascular disease ·Rural health
1 Introduction
Electrocardiogram (ECG) is a noninvasive diagnostic modality that has a substantial clinical impact on investigating the severity of cardiovascular diseases (CVD) [
1].
The ECG is an important screening tool that offers practitioners a wealth of informa-
tion that can be used alongside the history and clinical ﬁndings [2]. ECGs are pivotal
in the diagnosis of cardiac ischaemia and infarction which provides the evidence for pacemaker implantation, and detect inherited abnormalities such as cardiomyopathy
and long-QT syndrome [
3]. ECGs are also useful in detecting non-cardiac pathology,
for example, pulmonary emboli and electrolyte disorders. In accordance with the World Health Organization (WHO), India accounts for one-ﬁfth of CVD associ-
ated mortalities worldwide especially in younger population [
4]. Global Burden of
Disease study has also stated the age-standardized cardiovascular disease death rate of 272 per 100,000 population in India which is much higher than that of global
average of 235 [
5]. This pattern is uniform throughout the country despite wide vari-
ation in risk factors. India faces a great challenge in providing quality health care especially in rural domains due to lack of resources and trained healthcare providers.
The lack of resources for triaging or stratiﬁcation of patients has led to worsening of
the prognosis of patients. Moreover, the scarcity of super-specialized cardiologists
signiﬁcantly impacts the clinical prognosis of the patients intensifying the cardiovas-
cular disease burden. The hospital-based registries have neither been able to provide
accurate estimates of the CVD burden nor identify the disease drivers for the CVD
epidemic despite it being the largest cause of mortality. India is the world’s sixth-
largest country by land area and the world’s second-largest country by population
density. The country’s citizens come from a wide range of socioeconomic, linguistic,
cultural, and ethnic origins. The Indian subcontinent has historically served as a
conduit for several migratory waves originating in Africa, both on land and along
the coast [
6]. In mainland India, genetic research has shown four different ances-
tral populations, as well as a separate lineage in the Andaman and Nicobar Islands [
7]. As a melting pot of genetic variety, India also has strict inbreeding practices
and founder effects, resulting in the accumulation of adverse genetic variants [8].
India has a reported birth defect rate of 64.4 per 1000 live births [9]. Independent
research [10–13] has underlined India’s signiﬁcant genetic load. Because there was
no national newborn screening programme until recently, a large proportion of the

Earth Mover’s Distance-Based Automated … 5
Indian population was affected by genetic illnesses [14]. The genetic diversity of
India is well reﬂected in mitochondrial DNA (mtDNA), Y chromosomes, and candi-
date genes/markers, which have offered a reasonable knowledge of their related-
ness and divergence [
15–23]. Due to cultural and social traditions, consanguinity in
marriage is common in several subpopulations in India, resulting in the accumula- tion of genetic characteristics within groups [
7, 24]. According to the studies, there
is a signiﬁcant level of relatedness among subgroups, implying the accumulation
of detrimental variants [25, 26]. The Indian government took a national genome-
wide strategy to analyse population architecture and hunt for markers unique to the
Indian subcontinent. The Indian Genome Variation (IGV) group typed 900 genes
from approximately 1800 people across 55 endogamous populations using single
nucleotide polymorphisms (SNPs). The heterogeneity of subpopulations was high-
lighted by high heterozygosity values, varied allele frequencies, and frequent poly-
morphic haplotypes within the Indian community. In addition, novel mutations with
concurrent founder effects have been observed in the subcontinent [
27, 28]. The IGV
consortium’s discoveries have led to the discovery of markers and a better knowledge
of genotype–phenotype relationships across the Indian subcontinent. Susceptibility
or resistance to Plasmodium falciparum [29–33], risk of acquiring glaucoma [34],
homocysteine levels [35], risk of getting high-altitude pulmonary oedema [36, 37],
and various CVDs are only a few instances of phenotypically unique results of sub- population speciﬁc genotypes. As a result of the substantial genetic variation and
endogamous cultural practices, it is evident that genetic afﬁnities and differences between subpopulations must be deﬁned. These ﬁndings also highlight the genetic differences between Indian and other groups, highlighting the dangers of imputation
of genetic data from other populations. In clinical contexts, a generalization of the population architecture might obviously lead to erroneous conclusions. Many genetic variants that predispose people to CVDs have been discovered in the era of contempo-
rary genomics [
38]. Furthermore, single nucleotide polymorphisms (SNPs) in genes
that are associated with a favourable response to CVD medications have been antic- ipated by these genome-wide association analyses. SNPs associated with adverse
drug-induced toxicity in CVD patients have also been implicated in such investiga- tions. These ﬁndings, taken combined, provide a genetic foundation for stratifying individuals who are likely to beneﬁt from aggressive disease care by starting them
on pharmacological regimens tailored to their genetic variants. Cardiomyopathy is claimed to be more prevalent in the South Asian population than in the Western population [
39]. Polymorphisms in the genes that code for myosin-binding protein
C (MYBPC) and beta myosin heavy chain can cause cardiomyopathy by impairing the structural integrity and growth of the heart muscle. In the Indian population, only
a few studies have looked at the existence of polymorphisms in genes linked to CVD risk [
39–42]. Even fewer studies have included the Indian population, where CVD
mortality is said to be greater than in other parts of the nation (>300 deaths per 100,000
per year) [43]. The Indian demographics, being the most diverse, ethnicity holds a
strong position in determining the burden of CVDs. In considering Indian ethnic
identity and health, genetic basis may be relevant in two broad senses. First, the gene
pools of different ethnic groups of different sections of India may contain different

6
frequencies of alleles at some loci that are pertinent to health status or to disease
processes. Second, the phenotype consequent on a given genotype may vary between
ethnic groups because of interactions with environmental factors which covers all the
factors including prenatal effects, zonal nutritional inﬂuences, the preventive conse-
quences of health care, peer group pressures, educational level, religious instruction,
toxins in homes and in the air and water, occupational hazards, job stress, and expo-
sure to infectious agents and many others [
44]. Better understanding of the genetics
behind cardiovascular conditions will enhance the understanding of the pathophys- iological basis of these diseases, help identify novel pharmacological targets, as well as contribute to risk assessment, triaging as well as clinical management of
the patients. Also, recent data indicates that cardiac conduction measurements taken up by ECGs are heritable [
45–47] and have a genetic basis [48–51]. Insight into
which genes are implicated and the nature of the effect these genes have on the
heart, it is needed to include the advanced gene sequencing like Next Generation
Sequencing (NGS) into clinically actionable modalities. Development of novel arti-
ﬁcial intelligence (AI)-enabled ECG interpretation has become important in the clin-
ical ECG workﬂow since its inception over 50 years ago. It can serve as a crucial aide
to physicians generated through interpretation not only in resource-limited clinical
settings prevalent across the Indian subcontinent but also elsewhere across the world.
The availability of cheap, portable, accessible, and scalable computational platforms
with capabilities to process large-scale raw data will not only improve expert human
ECG interpretation by accurately triaging or prioritizing the most urgent conditions
but also importantly reduce the rates of misdiagnosed ECG interpretations. Hence,
an AI-enabled differentiation of normal from abnormal ECG using the categoriza-
tion based upon similarity matrices such as Earth Mover’s Distance (EMD) at the
community level which could consider the variations in genetic, regional, and ethnic
differences could play a role in modifying the graphs and hence the diagnosis. There
are no data available regarding the different variations in the healthy as well as
unhealthy ECG patterns at the community level. Our algorithm-enabled segregation
of normal records could make a pioneering step in differentiating it from abnormal
traces based on genetic background. In our work, we come out with novel frame-
work for computing the distance between images. The question of image similarity
is complex and delicate. It not only takes the corresponding bins into account, but
also considers the correlations between non-corresponding bins and thus is robust to
histogram shift and rotation.
2 Literature Review
ECG interpretation algorithms have gradually become the industry standard for
automating diagnosis and anatomical recognition on various systems throughout
the world. These algorithms use gender and age-speciﬁc criteria to provide a virtual
analysis for resting ECG interpretation, including detection of various cardiovascular
diseases and diagnostic aids to provide rhythm and morphology interpretation for a

Earth Mover’s Distance-Based Automated … 7
variety of patient populations, as well as an effective system for triaging patients to
prioritize the most serious conditions. Conventional statistical models, such as linear
and logistic regression, and even neural networks with few layers (often referred
to as “shallow models”) were previously constructed based on human-selected data
elements as inputs. In ECG analysis, for example, the inputs were morphological
and temporal parameters such as the QT, QRS, and RR intervals or QRS and/or
T-wave morphology, while the outputs were the ECG rhythm, serum potassium
level, and Left ventricular ejection fraction (LVEF). The model is constrained to
those attributes alone since the inputs are human-selected features. Furthermore,
any random or systematic error in computing features will be propagated to the
output, limiting the model’s accuracy. Deep learning allows the model to learn a
representation of the input data that includes features relevant to the task at hand,
without any human bias and without the need for human feature selection and engi-
neering, which can be time consuming, inaccurate, and reliant on expertise and
current physiological theories. The agnostic approach in a neural network is the best
representation, but it is also nonlinear, and the learned associations between input and
output data are currently unexplainable, making the model a “black box”—humans
cannot understand how the network makes decisions—which is one of the concerns
raised about deep-clinical learning’s application [
52]. Static and predictable algo-
rithms, such as Earth Mover’s Distance (EMDs), on the other hand, execute picture matching by computing perceptual similarity and give more meaningful and inter-
pretable matching answers. To reiterate, Artiﬁcial Neural Networks (ANNs) have a long and well-documented history of intrinsic instability, and their automated choices cannot be trusted to make life-saving decisions for a patient suffering from a severe
cardiac episode. Due to a lack of resources and competent workers, India confronts several obstacles in providing health care, particularly in rural areas. The fact that CVD-related fatalities account for 26% of all deaths in India demonstrates the enor-
mity of the public health challenge it faces [
53]. This is exacerbated by the lack of
appropriate algorithms to map out clinical resource allocation, particularly in rural underserved areas where doctor–patient ratios are low and access to inexpensive and
high-quality health care is severely limited. Despite the limits of current algorithms for effectively triaging individuals, the 12 lead ECG is still the most often used modality for identifying differences between normal ECGs as well as cardiac disor-
ders. Furthermore, the absence of well-structured niche-speciﬁc databases that span the various genetic foundations for referencing and analysis stymies the advancement of AI-assisted research to aid and optimize processes and clinical decision-making.
Aside from the complexity already discussed, the advancement of these technologies poses signiﬁcant technological, ethical, privacy, and therapeutic issues. Despite the
aforementioned challenges, AI-based automation of ECG analysis has a substan- tial socioeconomic impact and beneﬁt for LMICs like India. AI-based systems have a wide range of potential applications in health care, including screening, illness
diagnosis, patient risk assessment, and niche-speciﬁc best therapeutic options. The development of AI-enabled computational applications on these platforms would provide a valuable, precision public health tool for better management of the CVD

8
epidemic, alleviating a signiﬁcant burden on the government ﬁnances, given the
widespread use of mobile platforms and the Internet across the Indian subcontinent.
2.1 Earth Mover’s Distance
The formal deﬁnition of Earth Mover’s Distance is groundwork for “signatures” of
the form {(
1, 1), ( 2, 2), …, ( m, m)}, where xi is the centre of the data cluster
and pi is the number of points in the cluster. The two masses may be unequal if two
signatures are not normalized. Given two signatures
1, 1), …, ( m, m)} and
S
1, 1), …, ( n, n)}, the EMD is described relating to optimal ﬂow ij),
which curtails
W (R ) =
m
∑
i
n
∑
j
fij dij
where ij = i, j) represents dissimilarity between i and j. In the EMD explo-
ration,
The ﬂow (
ij) must satisfy the following constraints: fij ≥
n
∑
j
fij ≤ i ,
m
∑
i
fij ≤ j ,
m
∑
i
n
∑
j
fij =
⎛
⎝
m
∑
i
ri ,
n
∑
j
sj
⎞
⎠
The moment optimal ﬂow ij
* is determined; the Earth Mover’s Distance between
R
EMD(R ) =
⎛
⎝
m
∑
i
n
∑
j
f
∗
ij
dij
⎞
⎠/
⎛
⎝
m
∑
i
n
∑
j
fij )
⎞
⎠

Earth Mover’s Distance-Based Automated … 9
2.2 Various Distances
Assume
∑
represent the set of all the Borel subspace
of
can characterize fundamental distances and divergences between two distributions
Pr and Pg
(I)The total variation (TV) distance [54]
δ(
....
A
|Pr |.
(II) The Kullback–Leibler (KL) divergence
KL ∥Pg
.
log
Pr
Pg
Pr
where the two Pr and Pg are considered to be truly continuous and consequently
confess densities, as far as same estimate
2
. The KL divergence
[
54] is magniﬁcently unsymmetric and conceivable illimitable when there are
points so that Pg(
(III) The Jensen–Shannon (JS) divergence [54]
JS(Pr ) = (Pr ) + (Pg ),
where Pm is the average (Pr characterized because we selected
In Earth Mover’s Distance, the probability distributions converge, but in other
distances discussed above, the probability distributions diverge.
3 Statistics of ECG Training and Test Data Set
3.1 Data Set
This data set is collected from a medical camp conducted in Pulwama district, Jammu and Kashmir. Every ECG stream determines the electrical activity over one heartbeat.
Two categories of heartbeat were used, one is normal heartbeat and other is myocar- dial infarction. The data set consists of 25 training ECG images of both normal heartbeat and myocardial infarction of each category and ten testing images of both
normal and myocardial infarction of each category.

10
3.2 EMD-Based Analysis of Training Data
Firstly, we apply EMD on training data set. The training data set has 25 ECG images
in each of the two categories: one is normal heartbeat, and another is myocardial
infarction heartbeat. We keep these two categories of ECG images in two different
folders. Now the prime issue is that there are broad inequalities among healthy
ECG images and in unhealthy ECG images, and the delving query which is ECG
in the collection of healthy ECGs is the representative of normal ECGs which can
be exploited to measure amount of dissimilarity with test ECG images. Identical
query emerges to identify a representative unhealthy ECG which can be exploited
to measure dissimilarity with test ECGs. EMD technique answers these queries
positively, and EMD identiﬁes representative healthy ECG and unhealthy ECGs,
with pairwise EMD computation among healthy ECG collection and unhealthy ECG
collection, respectively.
Table 1 presents pairwise EMD computed over a sample of ﬁve ECG images so
that EMD matrix can be easily printed here. Complete EMD matrix for entire healthy
ECG data with dimension of 25
is available. In Table 1, the last row represents the sum of the EMDs of ﬁrst healthy
ECG in ﬁrst column with respect to other ECGs and similarly second column for second ECG and so on. Representative ECG is selected on the basis of minimum of
sum of the EMDs, in other words a representative ECG is picked which is closest to
all other healthy ECGs. A simple observation of sum of EMDs reveals that the third
column has minimum EMD, in other words third healthy ECG is closest to all other
healthy ECGs, so third ECG is considered as representative of all healthy ECGs.
Identical task is done for unhealthy ECGs to identify unhealthy representative
ECG, and its corresponding ﬁve-sample pairwise EMD matrix is shown in Table 2,
where the sum of EMDs entry contains sum of EMDs of ﬁrst ECG in ﬁrst column
Table 1
Healthy ECG
1
Healthy ECG 2 Healthy ECG 3 Healthy ECG 4 Healthy ECG 5
Healthy ECG 1 0.0000000
0.36726380.21245100.31083380.4157938
Healthy ECG 2 0.3647268
0.00000000.17768330.21618310.2967646
Healthy ECG 3 0.2110799
0.17928340.00000000.15943890.1562445
Healthy ECG 4 0.3129947
0.21957670.16048860.00000000.2897190
Healthy ECG 5 0.4157698
0.29624810.15724030.28895280.0000000
Sum of EMD’s 1.3045712
1.06237200.70786310.97540861.1585219

Earth Mover’s Distance-Based Automated … 11
Table 2
Unhealthy
ECG 1
Unhealthy ECG 2 Unhealthy ECG 3 Unhealthy ECG 4 Unhealthy ECG 5
Unhealthy ECG 1 0.0000000
0.31417700.23342880.22659010.4711726
Unhealthy ECG 2 0.3171718
0.00000000.28620490.20942080.3973091
Unhealthy ECG 3 0.2351154
0.28801500.00000000.33445080.4517883
Unhealthy ECG 4 0.2281418
0.21094360.33513630.00000000.3081703
Unhealthy ECG 5 0.4728687
0.39887570.45319630.30749640.0000000
Sum of EMD’s 1.253298
1.2120111.3079661.0779581.628440
with rest of all unhealthy ECGs, second column for second ECG, and so on. A simple
observation of sum of EMDs reveals that the fourth column has minimum EMD, in
other words fourth unhealthy ECG is closest to all other unhealthy ECGs, so fourth
ECG is considered as representative of all unhealthy ECGs. Identical task is carried
out on 25 unhealthy ECGs, and ECG number xx is selected as representative ECG
which is used to compare with rest of all unhealthy ECGs.
3.3 ECG Tagging with EMD
Test data set has 20 ECGs which includes both ten healthy and ten unhealthy ECGs, and these are compared against representative of healthy ECG. If test ECG is closer
to healthy ECG representative that is their EMD is smaller compared to unhealthy ECG representative, then the test ECG is tagged as “+1”, or if it is closer to unhealthy ECG representative, then it is tagged as “
overall accuracy of 60% and overall accuracy of 41.66%.
3.4 Algorithmic Framework Setup
EMD calculates Earth Mover’s Distance between distributions, and emd2d collates two distributions expressed as matrices. EMD feature of this task is done in R scripting
language (Fig.
1).

12
Initialization
Unhealthy ECG Training
Dataset
Healthy ECG Training
Dataset
Pairwise EMD
Calculation
Pairwise EMD
Calculation
Pairwise EMD
Calculation
Pairwise EMD
Calculation
Test ECG
Test ECG
H>U
Tag
Healthy
Tag
Unhealthy
Fig. 1

Earth Mover’s Distance-Based Automated … 13
Computational results
Test ECG ID
EMD with
respect to
healthy ECG
EMD with respect to unhealthy ECG EMD difference b/w healthy and unhealthy ECG Our disease tag Ground truthError
T1 0.163895
0.1991414 −+1+10
T20.1340776
0.1587326 −+1+10
T30.2082027
0.18104990.027153−+12
T40.1531404
0.1779477 −+1+10
T50.1307294
0.1722854 −+1+10
T60.2092571
0.18176790.027489−+12
T70.1531407
0.14858770.004553−+12
T80.3140732
0.2431530.07092−+12
T90.1515926
0.1835698 −+1+10
T100.1809756
0.190618 −+1+10
T110.1804128
0.1940299 −+1−−
T120.1562672
0.179345 −+1−−
T130.1536508
0.2079541 −+1−−
T140.1732075
0.17097120.002236−−0
T150.2482202
0.21119420.037026−−0
T160.2020969
0.219492 −+1−−
T170.1474309
0.181211 −+1−−
T180.1864928
0.2115686 −+1−−
T190.1650139
0.2058548 −+1−−
T200.1325707
0.146884 −+1−−
Computation of EMD for training healthy data
H1
0.1705611
0.2016369 −+1+10
H20.1438227
0.1609232 −+1+10
H30.1537007
0.1908258 −+1+10
H40.1456105
0.1852883 −+1+10
H50.1405678
0.1863929 −+1+10
H60.1845443
0.1616410.022903−+1+2
H70.171209
0.15682420.014385−+1+2
H80.1768019
0.1938189 −+1+10
H90.1644214
0.1686704 −+1+10
H100.1446915
0.1840659 −+1+10
H110.1251102
0.1712205 −+1+10
H120.1780795
0.15645950.02162−+1+2
(continued)

14
(continued)
Test ECG ID
EMD with
respect to
healthy ECG
EMD with respect to unhealthy ECGEMD difference b/w healthy and unhealthy ECGOur disease tagGround truthError
H13 0.1101195
0.177229 −+1+10
H140
0.2010695 −+1+10
H150.1948021
0.2117433 −+1+10
H160.1572228
0.2044841 −+1+10
H170.2382775
0.15222110.086056−+1+2
H180.1453631
0.2098166 −+1+10
H190.1533407
0.1819317 −+1+10
H200.2001808
0.16536520.034816−+1+2
Computation of EMD for training unhealthy data
UH1
0.1495372
0.1863064 −+1−−
UH20.1629056
0.2187011 −+1−−
UH30.1441234
0.1671735 −+1−−
UH40.1294056
0.1888588 −+1−−
UH50.1870512
0.1632910.02376−−0
UH60.1948595
0.2015528 −+1−−
UH70.1712404
0.1901085 −+1−−
UH80.1545726
0.1652415 −+1−−
UH90.2042541
00.204254−−0
UH100.1465899
0.1531914 −+1−−
UH110.1532719
0.1724425 −+1−−
UH120.1371117
0.1599319 −+1−−
UH130.1967969
0.2129231 −+1−−
UH140.1179075
0.1661596 −+1−−
UH150.1331991
0.1931969 −+1−−
UH160.1689143
0.1795368 −+1−−
UH170.1467736
0.1930697 −+1−−
UH180.1580702
0.1878384 −+1−−
UH190.1395352
0.1605549 −+1−−
UH200.1610765
0.2151979 −+1−−
Future implications
The Indian subcontinent’s current population is a mosaic of social, cultural, and
ethnic variety. Moreover, one billion people live on the Indian subcontinent today,

Earth Mover’s Distance-Based Automated … 15
representing hundreds of linguistic and ethnic groupings [54, 55]. Genetic and anthro-
pological research has revealed that the subcontinent’s peopling has a complicated
past, with contributions from several ancestral groups [56]. Recent autosomal single
nucleotide polymorphisms (SNP) and short tandem repeats (STR) investigations have
also revealed that Indian ethnic and linguistic groupings are genetically diverse [7–
57]. In 2010 research titled “Genetic diversity in India and the inference of Eurasian
population expansion,” Jinchuan Xing et al. found that the average number of single nucleotide polymorphisms (SNPs) per individual for Indians is 1.5, which is greater
than both European (1.3 SNPs) and East Asian individuals (1.1 SNPs). This ﬁnding
shows that Indian people have more genetic diversity than other Eurasian groups [
58].
Taken into this consideration, it can also be concluded that regarding this profound genetic diversity there must be variations in the healthy as well as unhealthy ECGs of the Indian individuals which is yet to explore. The genetic variation in the ECGs
can be very useful in classifying the ECGs into wider boundaries of CVDs and can provide better diagnosis of the various CVDs among the Indian population regard- less of the wise genetic variations. Our study was based on ECGs of a particular
community with a limited sample size which does disease tagging based on two classiﬁcations. But with larger and diverse ECGs, our software can be easily coupled
with conventional markers effective for the diagnostic tool for rapid screening of large populations in the community as currently no data accessible on the variances in ECG patterns at the community level throughout the Indian subcontinent, and this
repository would be a pioneering endeavour which can further classify ECGs into various CVDs tags with the help of EMD. Not only conﬁned to the CVDs, but our software can also be used in other disease tagging like lung cancer detection, brain
tumour classiﬁcation, etc. Our programme may easily be combined with traditional CVD indicators while also having a patient-agnostic genetic base, allowing it to be used as a quick screening tool for large populations in both community and tertiary
care settings. By incorporating large amount of genetic basis of the Indian patients, ECGs tagging will be much more resourceful in detecting CVDs not only conﬁned to zonal boundaries in India, instead it will be able to detect various CVDs/pre-CVDs
conditions inﬂuenced by genetic variations.
4 Conclusion
In this work, we used similarity detection metrics like EMD exploited more in image retrieval applications, and it can yield an accuracy of seventy-ﬁve per cent in
healthy ECG tagging. This work helps the patients make sense of ECGs and further recommend doctor consultation if required. This AI-enabled computational platform using EMD seamlessly will segregate the ECG image repository, and if the genetic
basis of the individual will be considered, then it provides accurate and precision- oriented triaging of ECG-based CVD anomalies even in resource-limited milieus to ensure precision-oriented diagnosis, thereby alleviating the clinical prognosis of the
patients especially in LMICs like India, where the scarcity of specialized cardiologists

16
signiﬁcantly affects the clinical prognosis of patients. Next Generation Sequencing
analysis is one of the anticipated strategies that can be implemented to map out the
frequency of novel molecular markers predictive of an impending cardiovascular
event at a community level after considering the Indian genetic, regional, and ethnic
differences that could play a statutory role in modifying the graphs and hence the
diagnosis. Currently, there is no data available regarding the different differences in
the ECG patterns at the community level across the Indian subcontinent, and this
repository would a pioneering effort with capabilities to be deployed. Our software
can be easily coupled with conventional markers of CVDs incorporating the patient-
agnostic genetic basis to enable its application as an effective diagnostic tool for rapid
screening of large populations in community settings as well as tertiary care settings.
As future work, this work can be extended to further classify ECGs into several
disease tags as with the ever-expanding number of variants associated with disease,
and we might expect that over time hundreds of gene regions will be implicated in
the pathophysiology of cardiac conditions.
Acknowledgements
support as grant in aid for the project entitled “Intelligent Decision Support System for Cardiovas-
cular Disease Detection Using ECG Traces at Tertiary Care Centres and Extended Community”
(ISRM/12(83)/2020 ID No. 2020-3500) to Dr. Rajiv Janardhanan.
Funding Indian Council of Medical Research, New Delhi, India.
References
1. Rundo F, Conoci S, Ortis A, Battiato S (2018) An advanced bio-inspired photoplethysmography (PPG) and ECG pattern recognition system for medical assessment. Sensors 18(2):405
2. Younker J (2011) Assessment of the cardiovascular system. Nursing the Cardiac Patient. Wiley- Blackwell, Chichester, pp 19–35
3. Jowett NI, Thompson DR (2007) Comprehensive coronary care. Elsevier Health Sciences
4. Garcia TB (2013) 12-lead ECG: The art of interpretation. Jones & Bartlett Publishers
5. Prabhakaran D, Jeemon P, Roy A (2016) Cardiovascular diseases in India: current epidemiology and future directions. Circulation 133(16):1605–1620
6. Reich D, Thangaraj K, Patterson N, Price AL, Singh L (2009) Reconstructing Indian population history. Nature 461(7263):489–494
7. Basu A, Sarkar-Roy N, Majumder PP (2016) Genomic reconstruction of the history of extant populations of India reveals ﬁve distinct ancestral components and a complex structure. Proc Natl Acad Sci 113(6):1594–1599
8. Nakatsuka N, Moorjani P, Rai N, Sarkar B, Tandon A, Patterson N, Bhavani GS, Girisha KM, Mustak MS, Srinivasan S, Kaushik A (2017) The promise of discovering population-speciﬁc disease-associated genes in South Asia. Nat Genet 49(9):1403–1407
9. Christianson A, Howson CP, Modell B (2005) March of Dimes: global report on birth defects, the hidden toll of dying and disabled children
10. Kaur A, Singh JR (2010) Chromosomal abnormalities: genetic disease burden in India. Int J Hum Genet 10(1–3):1–14
11. Singh I, Faruq M, Mukherjee O, Jain S, Pal PK, Srivastav MP, Behari M, Srivastava AK, Mukerji M (2010) North and South Indian populations share a common ancestral origin of Friedreich’s ataxia but vary in age of GAA repeat expansion. Ann Hum Genet 74(3):202–210

Earth Mover’s Distance-Based Automated … 17
12. Sachdeva K, Saxena R, Puri R, Bijarnia S, Kohli S, Verma IC (2012) Mutation analysis of
the CFTR gene in 225 children: identiﬁcation of ﬁve novel severe and seven reported severe
mutations. Genet Test Mol Biomarkers 16(7):798–801
13. Venugopal A, Chandran M, Eruppakotte N, Kizhakkillach S, Breezevilla SC, Vellingiri B (2018) Monogenic diseases in India. Mutation Res/Rev Mutation Res 776:23–31
14. Verma IC, Bijarnia-Mahay S, Jhingan G, Verma J (2015) Newborn screening: need of the hour in India. Indian J Pediatrics 82(1):61–70
15. Bamshad M, Kivisild T, Watkins WS, Dixon ME, Ricker CE, Rao BB, Naidu JM, Prasad BR, Reddy PG, Rasanayagam A, Papiha SS (2001) Genetic evidence on the origins of Indian caste populations. Genome Res 11(6):994–1004
16. Kivisild T, Rootsi S, Metspalu M, Mastana S, Kaldma K, Parik J, Metspalu E, Adojaan M, Tolk HV, Stepanov V, Gölge M (2003) The genetic heritage of the earliest settlers persists both in Indian tribal and caste populations. Am J Hum Genet 72(2):313–332
17. Das K, Malhotra KC, Mukherjee BN, Walter H, Majumder PP, Papiha SS (1996). Popula- tion structure and genetic differentiation among 16 tribal populations of central India. Hum Biol:679–705
18. Majumder PP, Roy B, Banerjee S, Chakraborty M, Dey B, Mukherjee N, Roy M, Thakurta
PG, Sil SK (1999) Human-speciﬁc insertion/deletion polymorphisms in Indian populations
and their possible evolutionary implications. Eur J Hum Genet 7(4):435–446
19. Thanseem I, Thangaraj K, Chaubey G, Singh VK, Bhaskar LV, Reddy BM, Reddy AG, Singh L (2006) Genetic afﬁnities among the lower castes and tribal groups of India: inference from Y chromosome and mitochondrial DNA. BMC Genet 7(1):1–11
20. Khan F, Pandey AK, Borkar M, Tripathi M, Talwar S, Bisen PS, Agrawal S (2008) Effect of sociocultural cleavage on genetic differentiation: a study from North India. Hum Biol 80(3):271–286
21. Borkar M, Ahmad F, Khan F, Agrawal S (2011) Paleolithic spread of Y-chromosomal lineage of tribes in eastern and northeastern India. Ann Hum Biol 38(6):736–746
22. Guha P, Das A, Dutta S, Bhattacharjee S, Chaudhuri TK (2015) Study of genetic diversity of KIR and TLR in the Rabhas, an endogamous primitive tribe of India. Hum Immunol 76(11):789–794
23. Chinniah R, Vijayan M, Thirunavukkarasu M, Mani D, Raju K, Ravi PM, Sivanadham R, Karuppiah B (2016) Polymorphic Alu insertion/deletion in different caste and tribal populations from South India. PLoS ONE 11(6):e0157468
24. Vadivelu MK (2016) Emergence of sociocultural norms restricting intermarriage in large social strata (endogamy) coincides with foreign invasions of India. Proc Natl Acad Sci 113(16):E2215–E2217
25. Bhasin MK, Nag S (2012) Consanguinity and its effects on fertility, mortality and morbidity in the Indian region: a review. Int J Hum Genet 12(4):197–301
26. Sengupta D, Choudhury A, Basu A, Ramsay M (2016) Population stratiﬁcation and under- representation of Indian subcontinent genetic diversity in the 1000 genomes project dataset. Genome Biol Evol 8(11):3460–3470
27. Indian Genome Variation Consortium+ 91-11-27667806+ 91-11-27667471 [email protected]. (2005) The Indian genome variation database (IGVdb): a project overview. Human Genet
118:1–11
28. Narang A, Roy RD, Chaurasia A, Mukhopadhyay A, Mukerji M, Dash D, Indian Genome Variation Consortium (2010) IGVBrowser–a genomic variation resource from diverse Indian populations. Database 2010
29. Sinha S, Mishra SK, Sharma S, Patibandla PK, Mallick PK, Sharma SK, Mohanty S, Pati SS, Mishra SK, Ramteke BK, Bhatt RM (2008) Polymorphisms of TNF-enhancer and gene for Fc population. Malar J 7(1):1–11
30. Sinha S, Qidwai T, Kanchan K, Anand P, Jha GN, Pati SS, Mohanty S, Mishra SK, Tyagi PK, Sharma SK, Venkatesh V (2008) Variations in host genes encoding adhesion molecules and susceptibility to falciparum malaria in India. Malar J 7(1):1–9

18
31. Sinha S, Arya V, Agarwal S, Habib S, Indian Genome Variation Consortium (2009) Genetic
differentiation of populations residing in areas of high malaria endemicity in India. J Genet
88(1):77–80
32. Jha P, Sinha S, Kanchan K, Qidwai T, Narang A, Singh PK, Pati SS, Mohanty S, Mishra SK, Sharma SK, Awasthi S, Indian Genome Variation Consortium (2012) Deletion of the APOBEC3B gene strongly impacts susceptibility to falciparum malaria. Infection, Genet Evol 12(1):142–148
33. Kanchan K, Pati SS, Mohanty S, Mishra SK, Sharma SK, Awasthi S, Venkatesh V, Habib S (2015) Polymorphisms in host genes encoding NOSII, C-reactive protein, and adhesion molecules thrombospondin and E-selectin are risk factors for Plasmodium falciparum malaria in India. Eur J Clin Microbiol Infect Dis 34(10):2029–2039
34. Bhattacharjee A, Banerjee D, Mookherjee S, Acharya M, Banerjee A, Ray A, Sen A, Ray K, & Indian Genome Variation Consortium (2008) Leu432Val polymorphism in CYP1B1 as a susceptible factor towards predisposition to primary open-angle glaucoma. Mole Vis 14:841
35. Kumar J, Garg G, Kumar A, Sundaramoorthy E, Sanapala KR, Ghosh S, Karthikeyan G,
Ramakrishnan L (2009). Single nucleotide polymorphisms in homocysteine metabolism
pathway genes: association of CHDH A119C and MTHFR C677T with hyperhomocys-
teinemia. Circ: Cardiovasc Genet 2(6):599–606
36. Aggarwal S, Negi S, Jha P, Singh PK, Stobdan T, Pasha MQ, Ghosh S, Agrawal A, Indian Genome Variation Consortium, Prasher B, Mukerji M (2010) EGLN1 involvement in high- altitude adaptation revealed through genetic analysis of extreme constitution types deﬁned in Ayurveda. Proc National Acad Sci 107(44):18961–18966
37. Aggarwal S, Gheware A, Agrawal A, Ghosh S, Prasher B, Mukerji M (2015) Combined genetic effects of EGLN1 and VWF modulate thrombotic outcome in hypoxia revealed by Ayurgenomics approach. J Transl Med 13(1):1–11
38. McPherson R (2013) From genome-wide association studies to functional genomics: new insights into cardiovascular disease. Can J Cardiol 29(1):23–29
39. Joshi P, Islam S, Pais P, Reddy S, Dorairaj P, Kazmi K, Pandey MR, Haque S, Mendis S, Rangarajan S, Yusuf S (2007) Risk factors for early myocardial infarction in South Asians compared with individuals in other countries. JAMA 297(3):286–294
40. Dhandapany PS, Sadayappan S, Xue Y, Powell GT, Rani DS, Nallari P, Rai TS, Khullar M, Soares P, Bahl A, Tharkan JM (2009) A common MYBPC3 (cardiac myosin binding protein C) variant associated with cardiomyopathies in South Asia. Nat Genet 41(2):187–191
41. Simonson TS, Zhang Y, Huff CD, Xing J, Watkins WS, Witherspoon DJ, Woodward SR, Jorde LB (2010) Limited distribution of a cardiomyopathy-associated variant in India. Ann Hum Genet 74(2):184–188
42. Reddy KS, Shah B, Varghese C, Ramadoss A (2005) Responding to the threat of chronic diseases in India. The Lancet 366(9498):1744–1749
43. Soman CR, Kutty VR, Safraj S, Vijayakumar K, Rajamohanan K, Ajayan K, PROLIFE Study Group (2011) All-cause mortality and cardiovascular mortality in Kerala State of India: results from a 5-year follow-up of 161 942 rural community dwelling adults. Asia Paciﬁc J Public Health 23(6):896–903
44. Bulatao RA, Anderson NB (2004) Understanding racial and ethnic differences in health in late life: a research agenda
45. Pilia G, Chen WM, Scuteri A, Orrú M, Albai G, Dei M, Lai S, Usala G, Lai M, Loi P, Mameli C (2006) Heritability of cardiovascular and personality traits in 6148 Sardinians. PLoS Genet 2(8):e132
46. Schrier L, Hadjipanayis A, Stiris T, Ross-Russell RI, Valiulis A, Turner MA, Zhao W, De Cock P, de Wildt SN, Allegaert K, van den Anker J (2020) Off-label use of medicines in neonates, infants, children, and adolescents: a joint policy statement by the European academy of Paedi- atrics and the European society for developmental perinatal and pediatric pharmacology. Eur J Pediatr 179(5):839–847
47. Newton-Cheh C, Guo CY, Wang TJ, O’donnell CJ, Levy D, Larson MG (2007) Genome-wide association study of electrocardiographic and heart rate variability traits: the Framingham heart study. BMC Med Genet 8(1):1–8

Earth Mover’s Distance-Based Automated … 19
48. Pfeufer A, Van Noord C, Marciante KD, Arking DE, Larson MG, Smith AV, Tarasov KV,
Müller M, Sotoodehnia N, Sinner MF, Verwoert GC (2010) Genome-wide association study
of PR interval. Nat Genet 42(2):153–159
49. Holm H, Gudbjartsson DF, Arnar DO, Thorleifsson G, Thorgeirsson G, Stefansdottir H, Gudjonsson SA, Jonasdottir A, Mathiesen EB, Njølstad I, Nyrnes A (2010) Several common variants modulate heart rate, PR interval and QRS duration. Nat Genet 42(2):117–122
50. Butler AM, Yin X, Evans DS, Nalls MA, Smith EN, Tanaka T, Li G, Buxbaum SG, Whitsel EA, Alonso A, Arking DE (2012). Novel loci associated with PR interval in a genome-wide association study of 10 African American cohorts. Circ: Cardiovasc Genet 5(6):639–646
51. Smith JG, Magnani JW, Palmer C, Meng YA, Soliman EZ, Musani SK, Kerr KF, Schnabel RB, Lubitz SA, Sotoodehnia N, Redline S, Candidate-gene Association Resource (CARe) Consor- tium (2011) Genome-wide association studies of the PR interval in African Americans. PLoS Genet 7(2):e1001304
52. Rudin C (2019) Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature Mach Intell 1(5):206–215
53. World Heart Federation. Fact sheet: cardiovascular diseases in India. https://world-heart-fed
eration.org/wpcontent/uploads/2017/05/Cardiovascular_diseases_in_India.pdf. Last accessed
13th May 2019
54. Singh KS (1992) People of India, vol 42. Anthropological Survey of India
55. Chaubey G, Metspalu M, Kivisild T, Villems R (2007) Peopling of South Asia: investigating the caste–tribe continuum in India. BioEssays 29(1):91–100
56. Majumder PP (2010) The human genetic history of South Asia. Curr Biol 20(4):R184–R187
57. Watkins WS, Thara R, Mowry BJ, Zhang Y, Witherspoon DJ, Tolpinrud W, Bamshad MJ, Tirupati S, Padmavati R, Smith H, Nancarrow D (2008) Genetic variation in South Indian castes: evidence from Y-chromosome, mitochondrial, and autosomal polymorphisms. BMC Genet 9(1):1–17
58. Xing J, Watkins WS, Hu Y, Huff CD, Sabo A, Muzny DM, Bamshad MJ, Gibbs RA, Jorde LB, Yu F (2010) Genetic diversity in India and the inference of Eurasian population expansion. Genome Biol 11(11):1–13

Cardial Disease Prediction
in Multi-variant Systems Using
MT-MrSBC Model
Pandiyan Nandakumar and Subhashini Narayan
Abstract
many reasons behind heart disease including smoking, heredity, and diabetes. Day
to day, people face various common symptoms of heart disease which are uncon-
sidered in a lethargic manner. This leads to serious and life-threatening complica-
tions. To predict these diseases in prior, several methods are existing which take in a
certain number of parameters for prognosis. The system proposed here is an ensemble
approach that combines the idea of the MT-MrSBC algorithm along with bagging
and boosting. The algorithm mentioned here overcomes the issues faced by other
algorithms in handling the multi-variant environment. The algorithm deploys itera-
tive techniques indulging bagging and boosting concepts that enhance the system.
The system trained is thus capable of predicting the disease of the patient. This helps
in taking precautionary measures by the patient which are life saving.
Keywords ·Prediction ·MT-MrSBC algorithm ·Cardiac disease
1 Introduction
Heart disease is the most pressing problem and a pivotal cause of death for both men and women. An enormous amount of ﬁndings is elaborated in medical associations (healing facility, medicinal focuses); still, this observation is not utilized properly.
The healthcare system is “data rich” however “knowledge poor”. Nowadays, heart disease was the main problem in most developed countries as well as for developing countries too. The rate of heart disease increases due to improper care and being
unaware of the condition. This kind of illness becomes life threatening when it is not monitored regularly. There are a few kinds of heart disease in this world, such as
P. Nandakumar (B) ·
School of Information Technology and Engineering, Vellore Institute of Technology,
Vellore 632014, Tamil Nadu, India
e-mail: [email protected]
S. Narayan e-mail:
[email protected]
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 H. K. Deva Sarma et al. (eds.),
Technology
https://doi.org/10.1007/978-981-19-5090-2_2
21

22 P. Nandakumar and S. Narayan
arrhythmia, coronary heart disease, congenital heart defects, cardiovascular disease,
and cardiomyopathy disease. Cardiovascular disease prompts different issues in the
human body such as sudden BP is raised and unexpected stroke. Cardiovascular
disease may also lead to the demise of the person in a short time duration.
As per recent statistics [1], it is found that due to heart diseases, 6.5 million
demises occur globally per annum. 0.9 million death rates (64%) are due to coronary heart disease and 0.4 million death rates (28%) by stroke. The death rate of 300 per
100,000 population in India is higher than the global average of 240 per 100,000 population. Adults born after 1970 are much more vulnerable to such deaths than those born earlier.
This paper is an initial attempt to reduce the time and efforts taken by the doctor by
systematizing the risk prognostic with the help of a concept called a binary classiﬁer. The imminence of cardiovascular disease is affected by environmental problems and
health services constraints. The heart attack increases due to various risk factors [
2]
including smoking, lack of exercise, high blood pressure, cholesterol, sugar levels, and pulse, an improper diet that leads to obesity, family history of coronary illness,
physical inertia, etc. Primal ﬁndings and treatment can keep heart disease from getting worse. Doctors will always depend on his or her knowledge and exposure rather than
an idea available in data that are covered in the database to make clinical decisions. Despite that, many doctors with their adept learning in all sub-profession are lacking resource. The information given by the patients may have repetitive symptoms which
leads them to suffer from varied diseases, and it may have matching symptoms. For this reason, the physicians cannot able to diagnose it validly since they are less experienced in handling it. Diagnosing heart disease perfectly at an earlier stage is
a very challenging task for many doctors because of the tough reliance on diverse factors.
Pertinent computer-based information and decision system can aid in achieving
the above test at a reduced cost. To prevent the increase in death rate, this algorithm is being used to predict the disease before undergoing the treatment on time and prevent itself from the severity of that disease. The term heart disease encompasses
the different kind of disease that affects the heart. The work done in this paper is as follows:
• In this paper, we present the experiment that was executed with the MT-MrSBC algorithm to build a predictive model for heart disease based on the available dataset.
• It also deploys the ensemble concepts of bagging and boosting that help to reduce the bias and error in prediction.
• The experimental result will show the better accuracy obtained using the proposed
model that achieves around 97% on three different datasets such as UCI Statlog,
Cleveland, and Framingham.
• It also compares with existing models such as logistic regression, K-nearest
neighbor, and decision tree.
The balance concept of this paper is organized as follows: In Sect. 2, related
works on the proposed model are surveyed. In Sects. 3 and 4, the proposed model is

Cardial Disease Prediction in Multi-variant … 23
explained with its architecture. In Sect. 5, the experimental results are shown in visual
representation. The future work is narrated in Sect. 6 that guides research toward
health sectors in disease classiﬁcation. Finally, in Sect. 7, the work is concluded with
the proposed model necessity in detecting heart disease earlier.
2 Related Works
The proposed method in this paper, MrSBC [3], explained the testing benchmark,
and results will be in an accurate manner. They explained about how it (MrSBC)
is efﬁcient in solving problems ﬁrst in an integrated approach, second in discrete
and continuous attributes, and third in the area of database schema by embedding
data model knowledge in generating rules. In the future, this model is useful for
comparison between other related data mining systems on a larger dataset.
The heart disease prediction [4] model is explained about how it was predicted
by advanced data mining techniques such as fuzzy logic. With the support from the member function, they had used fuzzy logic for execution comprising input variables,
rule-based, and output variables.
Ensemble learning [5] shows the combination of bagging and boosting methods
with six sub-classiﬁers. In each part, it used the concept of voting methodology for
predicting the ﬁnal result. The main objective is to integrate numerous puny learners
into an expediently more stable one. Based on the standardized data, there are two
notorious research areas in ensemble learning: They are ensemble clustering and
ensemble classiﬁcation. For clustering and classiﬁcation concepts [
6], heterogeneous
networks were used for a multi-type model with one proposed method called HENPC for predictive purposes. For bagging and boosting methods, Quinlan [
7] had applied
the above-said techniques and produced results in the decision tree learning and testing made in the given datasets. Based on the error rates value and the result, it shows that the boosted and bagged classiﬁers are notably lower than other classiﬁers
by using 18 datasets out of 27 available in bagging to C4.5.
By mathematical probability solution given in the following section, [8] explained
with Gaussian models providing the result in linear mean part alone. For further
classiﬁcation, in [9] the authors proposed a method based on the meta-path idea and
classiﬁed the objects collectively by describing them in a heterogeneous network.
A meta-path is a path consisting of an ordered link between two classiﬁed objects.
For predicting structured outputs [10] is applied by the way of learning models
with global and local tree ensembles employing different predicting tasks, such as hierarchical multi-label classiﬁcation, multi-target regression, and classiﬁcation. In
further work with some enhancement in global random forests, it can be useful in
feature ranking for structured outputs.
For more detail, in generating multiple versions of prediction, Breiman [11]
showed that the bagging method is performed well in predictive staging when applied to weak learners, and it provides how the accuracy is increased by doing changes

24 P. Nandakumar and S. Narayan
in the learning set. Classiﬁer type is used with the power of NB and K-NN algo-
rithms, namely called locally weighted Naïve Bayes classiﬁer. In this paper, they
have implemented a self-labeled weighted variant with Naïve Bayes as the base
classiﬁer and performed it with other semi-supervised classiﬁcation algorithms on
standard datasets. Finally, it concluded that the present technique had better accuracy
in most of the cases. Another technique is called self-labeling techniques which were
easy to implement and apply to almost all existing classiﬁers. These two points added
together with results found in [
12] used the Naïve Bayes model which guided us to
utilize the combination of bagging and boosting methods in the identical ensemble learning algorithm.
An extra idea for the issue we propose came from viable information by consid-
ering the prediction got at every cycle depending on various training sets. Each training set might zero in on various target attributes or on various properties of
the idea that should have been learned. Thus, we are normally ready to manage the multi-type classiﬁcation strategy, additionally by enhancing predictions of various types of objects types.
3 Methodology
In this part, we propose the disease prediction system related to cardiology which is based on an ensemble learning concept by MrSBC classiﬁers. Our proposed system is capable of predicting the disease of the patient with the help of the input, i.e.,
symptoms, given by them. To build a system that is capable of handling multi- variant data, we take in partially labeled heterogeneous data as our input and work iteratively on it. The ensemble approach is being richly exploited in our system.
The prior algorithm used here is multi-type multi-relational structural Naïve Bayes classiﬁer (MT-MrSBC). The proposed model description is shown in (Fig.
1). It is an
enhancement of the MrSBC algorithm that is proved to give better results due to its
iterative nature. Initially, the partially unlabeled heterogeneous data gathered is being
segregated into labeled and unlabeled separately. The purpose of the segregation
is to train the system with labeled data ﬁrst, followed by the unlabeled data. The
labeled data is being used to build a weak MrSBC model. This model cannot handle
diverse situations of the heterogenic environment. Hence, the above-said algorithm
is deployed along with the concepts of bagging and boosting using the unlabeled
data.
To express the concept of an ensemble in detail, the process of building and
combining numerous varied models to predict an outcome using a variety of modeling algorithms or training datasets. With the support of artiﬁcial intelligence where it is useful in an automatic diagnosis in healthcare sectors which is further subdivided into
machine learning that came into the picture with a lot of models focused on various human health-related problems. For more computational performance in diagnosing cardiac disease classiﬁcation types, many traditional machine learning methods are
used and trained with multiple models that use the identical learning algorithm to

Cardial Disease Prediction in Multi-variant … 25
Fig. 1
MT-MrSBC
produce an optimal solution. The ensembles take part in several groups of methods,
such as sequential and parallel which are called multi-classiﬁers, where a bunch of
hundreds or thousands of learners with a typical goal are combined to tackle the
speciﬁc issue.
The main cause of fault in learning comes from noise in data, bias, and variance.
The ensemble method helps it in limiting these variables. These strategies are chieﬂy intended to work on security and give the precision of machine learning models. The combinations of these numerous classiﬁers will bring about the decline in variance,
particularly in the space of unsound classiﬁers, and will create a more dependable classiﬁcation than a solitary classiﬁer.
The proposed model utilized the following dataset from the UCI repository such
as Cleveland, Statlog, and Framingham heart disease dataset. Cleveland dataset [13]
contains 76 attributes, but the most widely selected features are a subset of 14. In particular, most of the ML researchers used the Cleveland database. The UCI Statlog
(heart) dataset [
14] is a heart disease database that is the partially distinct form to a
database already present in the machine learning repository. The task used here is to detect the presence (1) or the absence (2) of heart disease. It has 270 instances with
13 attributes that have been used with no missing values. The UCI Framingham heart dataset [
15] was founded in the year 1948 under the U.S. Public Health Service and

26 P. Nandakumar and S. Narayan
was moved under the direct operations of the new national heart institute, NIH, in the
year 1949. It contains 4240 instances with 16 attributes. Participants were sampled
from Framingham, Massachusetts, that included both genders.
To use bagging and boosting concepts in the ensemble, a base learner algorithm is
required. The base learner algorithm used in our system is the MT-MrSBC algorithm. Bagging stands for bootstrap aggregation. Boosting refers to a group of algorithms that can convert weak learners to strong learners. Bagging and boosting are used to
get some N learners by producing enhanced data in the beginning phase. Then by the concept of random sampling, N new training datasets are created which has some changes from the original dataset. By sampling with new data, a few observations
might get rehashed in each new training dataset. In the case of bagging, any element with a similar probability has to appear in a new dataset. Nonetheless, the perceptions are scaled in boosting; consequently, some of them will partake in the new sets all
the more regularly.
Thus, we take in the labeled data from the heterogeneous data given and proceed
with ﬁrst-order classiﬁcation rules for developing a weak classiﬁer called MrSBC.
On developing the weak model of MrSBC, we ﬁnd that it is incapable of dealing
with the multi-variant nature of the system. The multi-variant nature of the system
here refers to the numerous symptoms of the patient which includes both primary
and secondary targets. The primary target indicates the common symptoms of the
patient. As cardiology is the mainstream here, common symptoms include chest
pain, sweating, etc. The secondary symptoms are the ones that add up to the decision
made by the system. It supplements the conﬁrmation of the decision taken by the
system. The major risk factors that can predict heart disease in an individual are high
blood pressure that leads to hypertension, high LDL cholesterol that is usually called
as bad cholesterol, low HDL cholesterol that is formerly called good cholesterol,
high triglycerides, high blood glucose level, family history of early heart disease,
smoking, and physical inactivity.
This is followed by training the system with unlabeled data. The unlabeled data is
a mixture of primary and secondary targets. As mentioned in [8], the random picking
of the target types makes the system viable to many circumstances and accustomed to it. Hence, we pick a random sample which may be either a primary or a secondary target type and train the system. By training the system, we mean that a classiﬁer is
built that can handle a similar situation if dealt with later in the system.
To build the classiﬁer, we use the Naïve Bayes theorem which can be stated as
P(C|A 1 A2 ... n ) = ( A 1 A2 ... n|C )P(C)/(A 1 A2 ... n) (1)
Given a record with some attributes 1, 2,…, n, the goal here is to predict class
C
For each given symptom in the target type taken, the probability is calculated
based on the theorem stated above. At this stage, the concept of bagging is deployed here. Bagging stands for bootstrapping and aggregating. Bootstrapping refers to the splitting up of the target type into multiple symptoms and ﬁnding the individual

Cardial Disease Prediction in Multi-variant … 27
probability of occurrence of the speciﬁc symptom. The average of those probabilities
is found out of which the highest is chosen. This is known as aggregation (Fig. 1).
While calculating the probabilities, there is a possibility for the occurrence of
zero probability which affects the system. This can be handled by using Laplacian smoothing stated as follows:
P i|C (N ic + )/(N c + ) (2)
where N is the number of possible values of the property predicate with which it is
associated and c is the distinct words in the dataset.
This helps to facilitate the inclusion of missing values such that it does not affect
the probabilities of other symptom occurrences. The classiﬁer built is being added to the labeled data to enhance the system in further prediction. The concept deployed
here is known as boosting. Thus, the ensemble concepts help in building the system further. On training the system, when the patient enters his symptoms as input to the system, it can make a prognosis of the disease that is more related to the symptoms
given by them. Due to the algorithm used, the system can cover a wide range of diseases that the patient may suffer. Hence, it is suitable to foresee the disease of the
patient before using our system.
4 System Architecture
Explanation
The proposed model architecture is shown in (Fig. 2). The above system undergoes
an initial training stage followed by a prediction stage. The admin trains the system
with partially labeled data using the MT-MrSBC algorithm with ensemble techniques
like bagging and boosting deployed. Though the system is initially a weak predictive
model, it transforms into a strong predictive model in subsequent iterations. Thus,
when the patient gives the symptoms he or she faced, as the input, the system predicts
the disease patient may have. The sequence in which the system works is depicted
above.
5 Experimental Approach
Let us check how the work explained above will result in practically with a simple example. Suppose we are building a classiﬁer that says the given symptoms lead to which of the given list of diseases. The different types of cardiac disease are shown
in (Fig.
3). Our training data has ten diseases that are given in Table 1 with their
relevant symptoms:

28 P. Nandakumar and S. Narayan
Fig. 2
Fig. 3
Now, what condition does S’s symptom belong to?
breath, fainting, stroke
Since Naive Bayes is a method for predicting probabilistic classiﬁer, we need
to calculate the probability the symptoms belong to pericarditis (D1), myocardial
infarction (D2), angina pectoris (D3), etc.,
It can be calculated as

Cardial Disease Prediction in Multi-variant … 29
Table 1
Disease
Symptoms Pericarditis Sharp pain in the left side of the chest, shortness of breath, heart
palpitation, heavy cough with fever, abdominal swelling, and
leg swelling Myocardial infarctionChest pain radiating to shoulder, arm, back, shortness of breath, cold sweat, fainting, hypertension, high blood cholesterol, smoking habit, diabetes Angina pectorisChest pain for less than 5 min, hypertension, clots in blood vessels ArrhythmiaHeartbeat does not work properly, fatigue or weakness, fainting Coronary arteryDecreasing blood ﬂow, chest pain, shortness of breath Dilated cardiomyopathyShortness of breath, fatigue, and swelling of the ankles, feet, and legs Hypertrophic cardiomyopathy Chest pain, dizziness, shortness of breath, fainting Mitral regurgitationAbnormal heart sound (heart murmur), shortness of breath (dyspnea), fatigue Mitral valve prolapseFluttering or rapid heartbeat called palpitations, shortness of breath especially with exercise, dizziness, passing out or fainting known as syncope, panic, and anxiety Pulmonary stenosisHeart murmur, fatigue, shortness of breath, especially during exertion, chest pain, and loss of consciousness
P(D1 i ) = [P(D1 i ) ∗ (D1)]P(S i ) (3)
P(D2 i ) = [P(D2 i ) ∗ (D2)]P(S i ) (4)
P(D3 i ) = [P(D3 i ) ∗ (D3)]P(S i ) (5)
P(Si /) = (CP) ∗ (SoB) ∗ (F) ∗ (S) (6)
where CP—chest pain, SoB—shortness of breath,
From Fig. 4, out of the ten diseases, D2 (myocardial infarction) has the highest
probability. Then, the given symptoms refer to myocardial infarction. We can
conclude that the patient suffering from above given symptoms (S) (from Fig. 4
and Table 2) suffers from myocardial infarction. This gives a glimpse of how the
system works in training the system using the Naïve Bayes theorem.
The comparative analysis of the proposed model with existing models is given
in Table 3. It exhibits that the earlier machine learning models achieved an accu-
racy of around 78% to 95% with the different datasets by various researchers. The proposed model uses the concept of ensemble approach with MT-MrSBC which is
an extension of the Naïve Bayes classiﬁcation method to the multi-relational setting

30 P. Nandakumar and S. Narayan
1.2031.101 1.0451.011
Pericarditis
Myocardial
Infarction
Angina
Pectoris
Arrhythmia
Coronary
artery
Dilated
cardiomyopa
thy
Hypertrophic
cardiomyopa
thy
Mitral
regurgitation
Mitral valve
prolapse
Pulmonary
stenosis
123456789
1.256 1.2111.3611.458 1.345 1.321
10
0
0.5
1
1.5
2

Type of Diseases
Fig. 4
Table 2
suffering from a different
kind of cardiac disease
S. No.
Disease
Va l u eResult
1 Pericarditis
1.361Medium
2Myocardial infarction
1.458High
3Angina pectoris
1.203Medium
4Arrhythmia
1.101Medium
5Coronary artery
1.256Medium
6Dilated cardiomyopathy
1.345Medium
7Hypertrophic cardiomyopathy
1.211Medium
8Mitral regurgitation
1.045Medium
9Mitral valve prolapses
1.011Medium
10Pulmonary stenosis
1.321Medium
and has achieved an accuracy of around 97%. The model execution starts with the
splitting of the training and testing datasets as labeled and unlabeled data. Based
on the symptoms or features, we have passed the valid input data to the proposed
ensemble model that classiﬁes the ﬁnal prediction of heart disease. Figure
5 shows
the visual representation of the proposed model performance. The results show that
the proposed MT-MrSBC achieves better accuracy than the earlier research.
6 Future Works
Nowadays, the health-related issue is a disaster for all human beings. For predicting, analyzing, and treatment purposes, it needs so much cost, as well as for data handling
process, and is difﬁcult. This system helps in predicting the disease of the patient according to the symptoms given by them. This work can be further enhanced by adding various features that include the suggestion of the various hospitals and avail-
able doctors in those hospitals. Patients can choose doctors at their convenience. Also,

Cardial Disease Prediction in Multi-variant … 31
Table 3
References
Year
Models DatasetAccuracy (%)
Tay et al. [16] 2015
Support vector
machine (SVM)
Cardiovascular health study (CHS) dataset 95.3, 84.8, and 90.1
Mufudza and Erol [
17]
2016
Poisson mixture regression model Cleveland clinic foundation heart disease data set 86.7
Dogan et al. [18]2018
Random forest classiﬁcation model for symptomatic CHD and integrated genetic-epigenetic algorithms Framingham heart study 78
Dwivedi [19]2018
Artiﬁcial neural network (ANN), support vector machine (SVM),
logistic regression, K-nearest neighbor (K-NN), classiﬁcation tree, and Naïve Bayes
Statlog heart disease dataset 85
Haq et al. [20]2018
K-NN, ANN, SVM, DT, and NB Cleveland heart disease dataset 2016 83–89
Mohan et al. [21]2019
Hybrid random forest with a linear model (HRFLM) Cleveland dataset88.7
Javeed et al. [22]2019
Random search algorithm (RSA), random forest model, grid search algorithm Cleveland dataset93.33
Vankara and Devi [
23]
2020
Ensemble learning by cuckoo search UCI83–86
Proposed model: MT-MrSBC 2021
Multi-type multi-relational structural Naïve Bayes classiﬁer UCI—Cleveland, Statlog, and Framingham dataset 97.3
they can choose nearby hospitals from a stack of hospitals and select the doctors avail-
able in the cardiology department. The patient will be able to treat his disease in a
nearby available hospital which saves life and time. In real time, the patient has to
wait for the appointment, whereas doing it online saves time in emergencies. To
enhance the system, it can also provide appointments online with the chosen doctor,
and also patients could see the available doctors on that particular day or week. The

32 P. Nandakumar and S. Narayan
Fig. 5
patient will also be able to ﬁx an appointment and look for availability of doctors
and be provided with temporary diagnosis to be aware of right on the spot within
a few minutes. In the future, our model can be enhanced by applying feature selec-
tion concepts that are not used in existing types. The major problem in machine
learning techniques is the high dimensionality from the datasets that leads to an
analysis of many features which results in a large amount of storage requirements.
For this reason, in the future, the proposed model can be further enhanced with the
bio-inspired algorithm for producing an optimal result.
7 Conclusion
Heart attack is a crucial health problem in human society. It is the major health challenge in the globe for the last decade. Nowadays, many research works have been
carried out to predict, detect, and diagnose heart disease, which could prove to be vital in combating the disease. Since it is creating a maximum portion of the information that medicos are incapable to deduce and use expertly, machine learning with an
ensemble model has transpired as a more errorless and potent model in a voluminous kind of medical obstacles such as prognosis, prophecy, and invasion. This paper results facilitate taking precautionary measures to control heart disease. The possible
advantages of applying machine learning strategies with a reasonable calculation will lessen the illness rates and passing disappointments of the worldwide populace. The models here are trained and validated against a sample test dataset. The Naïve
Bayes algorithm is the most efﬁcient model to predict patients with heart disease. This paper has summarized the state for predicting cardiac disease using multi-type
multi-relational structural Naïve Bayes classiﬁer (MT-MrSBC) concerning ease of

Cardial Disease Prediction in Multi-variant … 33
model interpretation and accuracy. In our work, good results with improved accuracy
were given in the prediction of cardiac diseases.
Acknowledgements
Funding No funding for this research work.
Compliance with Ethical Standards
participants performed by any of the authors.
References
1. Virani SS et al (2020) Heart disease and stroke statistics—2020 update: a report from the American heart association. Circulation 141(9):e139–e596
2. Psaltopoulou T et al (2017) Socioeconomic status and risk factors for cardiovascular disease: impact of dietary mediators. Hellenic J Cardiol 58(1):32–42
3. Ceci M, Appice A, Malerba D (2003) Mr-SBC: a multi-relational Naive Bayes classiﬁer. In: PKDD. Springer, pp 95–106
4. Kowsigan M et al (2017) Heart disease prediction by analysing various parameters using fuzzy logic. Pak J Biotechnol 14(2):157–161
5. Kotsiantis S (2011) Combining bagging, boosting, rotation forest and random subspace methods. Artif Intell Rev 35(3):223–240
6. Pio G et al (2018) Multi-type clustering and classiﬁcation from heterogeneous networks. Inf Sci 425(2018):107–126
7. Quinlan JR (1996) Bagging, boosting, and c4. 5. In: AAAI/IAAI, vol 1, pp 725–730
8. Roberts GO, Sahu SK (1997) Updating schemes, correlation structure, blocking and parame- terization for the Gibbs sampler. J R Statis Soc: Ser B (Statis Methodol) 59(2):291–317
9. Kong X et al (2012) Meta path-based collective classiﬁcation in heterogeneous information networks. In: Proceedings of the 21st ACM international conference on Information and
knowledge management, pp 1567–1571
10. Kocev D et al (2013) Tree ensembles for predicting structured outputs. Patt Recogn 46(3):817– 833
11. Breiman L (1996) Bagging predictors. Mach Learn 24(2):123–140
12. Karlos S et al (2017) Locally application of Naive Bayes for self-training. Evol Syst 8(1):3–18
13. Detrano R et al (1989) International application of a new probability algorithm for the diagnosis of coronary artery disease. The Am J Cardiol 64(5):304–310
14. Xiao L, Wang X et al (2017) A hybrid classiﬁcation system for heart disease diagnosis based on the RFRS method. Comput Math Meth Med 2017:11. Article ID 8272091.
https://doi.org/
10.1155/2017/8272091
15. Splansky GL, Corey D et al (2007) The third generation cohort of the national heart, lung, and blood institute’s Framingham heart study: design, recruitment, and initial examination. Am J Epidemiol 165(11):1328–1335.
https://doi.org/10.1093/aje/kwm021 Epub 2007 Mar 19
PMID: 17372189
16. Tay D et al (2014) The effect of sample age and prediction resolution on myocardial infarction risk prediction. IEEE J Biomed Health Inform 19(3):1178–1185
17. Mufudza C, Erol H (2016) Poisson mixture regression models for heart disease prediction. Comput Math Meth Med
18. Dogan MV et al (2018) Integrated genetic and epigenetic prediction of coronary heart disease in the Framingham heart study. PloS One 13(1):e0190549

34 P. Nandakumar and S. Narayan
19. Dwivedi AK (2018) Performance evaluation of different machine learning techniques for
prediction of heart disease. Neural Comput Appl 29(10):685–693
20. Haq AUl et al (2018) A hybrid intelligent system framework for the prediction of heart disease using machine learning algorithms. Mobile Inf Syst
21. Mohan S, Thirumalai C, Srivastava G (2019) Effective heart disease prediction using hybrid machine learning techniques. IEEE Access 7:81542–81554
22. Javeed A et al (2019) An intelligent learning system based on random search algorithm and optimized random forest model for improved heart disease detection. IEEE Access 7(2019):180235–180243
23. Vankara J, Devi GL (2020) PAELC: predictive analysis by ensemble learning and classiﬁcation heart disease detection using beat sound. Int J Speech Technol 23(1):31–43

Random documents with unrelated
content Scribd suggests to you:

The Project Gutenberg eBook of The Romany
Rye

This ebook is for the use of anyone anywhere in the United
States and most other parts of the world at no cost and with
almost no restrictions whatsoever. You may copy it, give it away
or re-use it under the terms of the Project Gutenberg License
included with this ebook or online at www.gutenberg.org. If you
are not located in the United States, you will have to check the
laws of the country where you are located before using this
eBook.
Title: The Romany Rye
Author: George Borrow
Editor: William I. Knapp
Release date: January 24, 2017 [eBook #54048]
Most recently updated: August 9, 2019
Language: English
Credits: Transcribed from the 1907 John Murray edition by David
Price
*** START OF THE PROJECT GUTENBERG EBOOK THE ROMANY RYE
***

There are several editions of this ebook in the Project
Gutenberg collection. Various characteristics of each ebook
are listed to aid in selecting the preferred file.
Click on any of the filenumbers below to quickly view each
ebook.
21206(Plain HTML file)
25071(Plain HTML file)
422(Plain HTML file)
54048(Illustrated HTML file)
Transcribed from the 1907 John Murray edition by David Price, email
[email protected]

THE ROMANY RYE
A SEQUEL TO “LAVENGRO”

BY GEORGE BORROW

A NEW EDITION CONTAINING THE UNALTERED
TEXT OF THE ORIGINAL ISSUE, WITH
NOTES. ETC., BY THE AUTHOR OF
“THE LIFE OF GEORGE BORROW”

LONDON
JOHN MURRAY, ALBEMARLE STREET
1907
First Edition 1857
Second Edition 1858
Third Edition 1872
Fourth Edition 1888
Fifth Edition 1896
Siñth (Definitive ) Edition 6/-March, 1900
Reprinted June, 1903
Reprinted Thin PaperAug., 1905
Reprinted 6/- Oct., 1906
Reprinted Sept., 1907

Reprinted 2/6 netSept., 1907

ADVERTISEMENT.
(1857.)
It having been frequently stated in print that the book called
Lavengro was got up expressly against the popish agitation in the
years 1850–51, the author takes this opportunity of saying that the
principal part of that book was written in the year ’43, that the whole
of it was completed before the termination of the year ’46, and that it
was in the hands of the publisher in the year ’48.
[0a]
And here he
cannot forbear observing, that it was the duty of that publisher to
have rebutted a statement which he knew to be a calumny; and also
to have set the public right on another point dealt with in the
Appendix to the present work, more especially as he was the
proprietor of a Review enjoying, however undeservedly, a certain sale
and reputation.
But take your own part, boy!
For if you don’t, no one will take it for you.
With respect to Lavengro, the author feels that he has no reason to
be ashamed of it. In writing that book he did his duty, by pointing
out to his country-people the nonsense which, to the greater part of
them, is as the breath of their nostrils, and which, if indulged in, as it
probably will be, to the same extent as hitherto, will, within a very
few years, bring the land which he most loves beneath a foreign yoke
—he does not here allude to the yoke of Rome.
Instead of being ashamed, has he not rather cause to be proud of a
book which has had the honour of being rancorously abused and

execrated by the very people of whom the country has least reason
to be proud?
[0b]

“One day Cogia Efendy went to a bridal festival. The masters of
the feast, observing his old and coarse apparel, paid him no
consideration whatever. The Cogia saw that he had no chance of
notice; so going out, he hurried to his house, and, putting on a
splendid pelisse, returned to the place of festival. No sooner did
he enter the door than the masters advanced to meet him, and
saying, ‘Welcome, Cogia Efendy,’ with all imaginable honour and
reverence, placed him at the head of the table, and said, ‘Please
to eat, Lord Cogia’. Forthwith the Cogia, taking hold of one of
the furs of his pelisse, said, ‘Welcome my pelisse; please to eat,
my lord’. The masters looking at the Cogia with great surprise,
said, ‘What are you about?’ Wher eupon the Cogia replied, ‘As it
is quite evident that all the honour paid, is paid to my pelisse, I
think it ought to have some food too’.”—Pleasantries of the Cogia
Nasr Eddin Efendi.

CONTENTS.
PAGE
CHAPTER I.
The Making of the Linch-pin—The Sound Sleeper—Breakfast—
The Postillion’s Departure
1
CHAPTER II.
The Man in Black—The Emperor of Germany—Nepotism—Donna
Olympia—Omnipotence—Camillo Astalli—The Five Propositions
5
CHAPTER III.
Necessity of Religion—The Great Indian One—Image-worship—
Shakespeare—The Pat Answer—Krishna—Amen
9
CHAPTER IV.
The Proposal—The Scotch Novel—Latitude—Miracles—Pestilent
Heretics—Old Fraser—Wonderful Texts—No Armenian
16
CHAPTER V.
Fresh Arrivals—Pitching the Tent—Certificated Wife—High-flying
Notions
28
CHAPTER VI.
The Promised Visit—Roman Fashion—Wizard and Witch—
Catching at Words—The Two Females—Dressing of Hair—The
New Roads—Belle’s Altered Appearance—Herself Again
32
CHAPTER VII.

The Festival—The Gypsy Song—Piramus of Rome—The
Scotchman—Gypsy Names
40
CHAPTER VIII.
The Church—The Aristocratical Pew—Days of Yore—The
Clergyman—“In what would a Man be Profited?”
48
CHAPTER IX.
Return from Church—The Cuckoo and Gypsy—Spiritual Discourse53
CHAPTER X.
Sunday Evening—Ursula—Action at Law—Meridiana—Married
Already
60
CHAPTER XI.
Ursula’s Tale—The Patteran—The Deep Water—Second Husband72
CHAPTER XII.
The Dingle at Night—The Two Sides of the Question—Roman
Females—Filling the Kettle—The Dream—The Tall Figure
78
CHAPTER XIII.
Visit to the Landlord—His Mortifications—Hunter and his Clan—
Resolution
86
CHAPTER XIV.
Preparations for the Fair—The Last Lesson—The Verb Siriel 89
CHAPTER XV.
The Dawn of Day—The Last Farewell—Departure for the Fair—
The Fine Horse—Return to the Dingle—No Isopel
95
CHAPTER XVI.
Gloomy Forebodings—The Postman’s Mother—The Letter—Bears
and Barons—The Best of Advice
99
CHAPTER XVII.

The Public-house—Landlord on His Legs Again—A Blow in
Season—The Way of the World—The Grateful Mind—The Horse’s
Neigh
106
CHAPTER XVIII.
Mr. Petulengro’s Device—The Leathern Purse—Consent to
Purchase a Horse
113
CHAPTER XIX.
Trying the Horse—The Feats of Tawno—Man with the Red
Waistcoat—Disposal of Property
117
CHAPTER XX.
Farewell to the Romans—The Landlord and his Niece—Set out as
a Traveller
122
CHAPTER XXI.
An Adventure on the Roads—The Six Flint Stones—A Rural Scene
—Mead—The Old Man and his Bees
124
CHAPTER XXII.
The Singular Noise—Sleeping in a Meadow—The Book—Cure for
Wakefulness—Literary Tea Party—Poor Byron
131
CHAPTER XXIII.
Drivers and Front Outside Passengers—Fatigue of Body and Mind
—Unexpected Greeting—My Inn—The Governor—Engagement
136
CHAPTER XXIV.
An Inn of Times gone by—A First-rate Publican—Hay and Corn—
Old-fashioned Ostler—Highwaymen—Mounted Police—Grooming
140
CHAPTER XXV.
Stable Hartshorn—How to Manage a Horse on a Journey—Your
Best Friend
145
CHAPTER XXVI.

The Stage-coachmen of England—A Bully Served Out—
Broughton’s Guard—The Brasen Head
150
CHAPTER XXVII.
Francis Ardry—His Misfortunes—Dog and Lion Fight—Great Men
of the World
158
CHAPTER XXVIII.
Mr. Platitude and the Man in Black—The Postillion’s Adventures—
The Lone House—A Goodly Assemblage
163
CHAPTER XXIX.
Deliberations with Self—Resolution—Invitation to Dinner—The
Commercial Traveller—The Landlord’s Offer—The Comet Wine
170
CHAPTER XXX.
Triumphal Departure—No Season like Youth—Extreme Old Age—
Beautiful England—The Ratcatcher—A Misadventure
175
CHAPTER XXXI.
Novel Situation—The Elderly Individual—The Surgeon—A Kind
Offer—Chimerical Ideas—Strange Dream
179
CHAPTER XXXII.
The Morning after a Fall—The Teapot—Unpretending Hospitality
—The Chinese Student
185
CHAPTER XXXIII.
Convalescence—The Surgeon’s Bill—Letter of Recommendation—
Commencement of the Old Man’s History
191
CHAPTER XXXIV.
The Old Man’s Story continued—Misery in the Head—The
Strange Marks—Tea-dealer from London—Difficulties of the
Chinese Language
201
CHAPTER XXXV.

The Leave-taking—Spirit of the Hearth—What’s o’Clock 209
CHAPTER XXXVI.
Arrival at Horncastle—The Inn and Ostlers—The Garret—The
Figure of a Man with a Candle
211
CHAPTER XXXVII.
Horncastle Fair 214
CHAPTER XXXVIII.
High Dutch 221
CHAPTER XXXIX.
The Hungarian 223
CHAPTER XL.
The Horncastle Welcome—Tzernebock and Bielebock 238
CHAPTER XLI.
The Jockey’s Tale—Thieves’ Latin—Liberties with Coin—The
Smasher in Prison—Old Fulcher—Every one has his Gift—Fashion
of the English
244
CHAPTER XLII.
A Short-tempered Person—Gravitation—The Best Endowment—
Mary Fulcher—Fair Dealing—Horse-witchery—Darius and his
Groom—The Jockey’s Tricks—The Two Characters—The Jockey’s
Song
258
CHAPTER XLIII.
The Church 273
CHAPTER XLIV.
An Old Acquaintance 276
CHAPTER XLV.
Murtagh’s Tale 283

CHAPTER XLVI.
Murtagh’s Story continued—The Priest, Exorcist, and Thimble-
engro—How to Check a Rebellion
290
CHAPTER XLVII.
Departure from Horncastle—Recruiting Sergeant—Kauloes and
Lolloes
300

APPENDIX.
CHAPTER I.
A Word for Lavengro 302
CHAPTER II.
On Priestcraft 310
CHAPTER III.
On Foreign Nonsense 317
CHAPTER IV.
On Gentility Nonsense—Illustrations of Gentility 320
CHAPTER V.
Subject of Gentility continued 323
CHAPTER VI.
On Scotch Gentility Nonsense—Charlie o’er the Waterism 334
CHAPTER VII.
Same subject continued 341
CHAPTER VIII.
On Canting Nonsense 346
CHAPTER IX.
Pseudo-Critics 354

CHAPTER X.
Pseudo-Radicals 362
CHAPTER XI.
The Old Radical 368

Editor’s Notes 379
Gypsy List 389
Bibliography 393

LIST OF ILLUSTRATIONS.
East Dereham, Norfolk (referred to as “Pretty D—,”
George Borrow’s Birthplace) (photogravure)
Frontispiece
The Old Church, St. Giles, at Willenhall, Staffordshire
(rebuilt 1867)
To face page
48
Porch of St. Nicholas Church, East Dereham 50
The Old “Bull’s Head,” Wolverhampton Street,
Willenhall
106
The “Swan” Inn, Stafford (“My Inn—a very large
Building with an Archway”)
136
High Street, Horncastle 215
The Horse Fair, Horncastle 220
Horncastle Church in 1820 (since restored) 273

CHAPTER I.
I awoke at the first break of day, and, leaving the postillion fast
asleep, stepped out of the tent. The dingle w as dank and dripping. I
lighted a fire of coals, and got my forge in readiness. I then
ascended to the field, where the chaise was standing as we had left it
on the previous evening. After looking at the cloud-stone near it,
now cold, and split into three pieces, I set about prying narrowly into
the condition of the wheel and axle-tree. The latter had sustained no
damage of any consequence, and the wheel, as far as I was able to
judge, was sound, being only slightly injured in the box. The only
thing requisite to set the chaise in a travelling condition appeared to
be a linch-pin, which I determined to make. Going to the companion
wheel, I took out the linch-pin, which I carried down with me to the
dingle, to serve me as a model.
I found Belle by this time dressed, and seated near the forge. With a
slight nod to her like that which a person gives who happens to see
an acquaintance when his mind is occupied with important business,
I forthwith set about my work. S electing a piece of iron which I
thought would serve my purpose, I placed it in the fire, and plying
the bellows in a furious manner, soon made it hot; then seizing it with
the tongs, I laid it on the anvil, and began to beat it with my
hammer, according to the rules of my art. The dingle resounded with
my strokes. Belle sat still, and occasionally smiled, but suddenly
started up, and retreated towards her encampment, on a spark which
I purposely sent in her direction alighting on her knee. I f ound the
making of a linch-pin no easy matter; it was, however, less difficult
than the fabrication of a pony-shoe; my work, indeed, was much
facilitated by my having another pin to look at. I n about three-

quarters of an hour I had succeeded tolerably well, and had produced
a linch-pin which I thought would serve. During all this time,
notwithstanding the noise which I was making, the postillion never
showed his face. His non-appearance at first alarmed me: I was
afraid he might be dead, but, on looking into the tent, I found him
still buried in the soundest sleep. “He must surely be descended
from one of the seven sleepers,” said I, as I turned away, and
resumed my work. My work finished, I took a little oil, leather and
sand, and polished the pin as well as I could; then, summoning Belle,
we both went to the chaise, where, with her assistance, I put on the
wheel. The l inch-pin which I had made fitted its place very well, and
having replaced the other, I gazed at the chaise for some time with
my heart full of that satisfaction which results from the consciousness
of having achieved a great action; then, after looking at Belle in the
hope of obtaining a compliment from her lips, which did not come, I
returned to the dingle, without saying a word, followed by her. Belle
set about making preparations for breakfast; and I taking the kettle
went and filled it at the spring. Ha ving hung it over the fire, I went
to the tent in which the postillion was still sleeping, and called upon
him to arise. He a woke with a start, and stared around him at first
with the utmost surprise, not unmixed, I could observe, with a
certain degree of fear. At last, looking in my face, he appeared to
recollect himself. “I had quite forgot,” said he, as he got up, “where I
was, and all that happened yesterday. However, I remember now the
whole affair, thunder-storm, thunder-bolt, frightened horses, and all
your kindness. Come, I must see after m y coach and horses; I hope
we shall be able to repair the damage.” “The damage is already quite
repaired,” said I, “as you will see, if you come to the field above.”
“You don’t say so,” said the postillion, coming out of the tent; “well, I
am mightily beholden to you. Good-morning, young gentlewoman,”
said he, addressing Belle, who, having finished her preparations was
seated near the fire. “Good morning, young man,” said Belle, “I
suppose you would be glad of some breakfast; however, you must
wait a little, the kettle does not boil.” “Come and look at your
chaise,” said I; “but tell me how it happened that the noise which I
have been making did not awake you; for three-quarters of an hour

at least I was hammering close at your ear.” “I heard you all the
time,” said the postillion, “but your hammering made me sleep all the
sounder; I am used to hear hammering in my morning sleep. There’s
a forge close by the room where I sleep when I’m at home, at my
inn; for we have all kinds of conveniences at my inn—forge,
carpenter’s shop, and wheelwright’s—so that when I heard you
hammering I thought, no doubt, that it was the old noise, and that I
was comfortable in my bed at my own inn.” We now ascended to the
field, where I showed the postillion his chaise. He look ed at the pin
attentively, rubbed his hands, and gave a loud laugh. “I s it not well
done?” said I. “It wi ll do till I get home,” he replied. “And that is all
you have to say?” I demanded. “ And that’s a good deal,” said he,
“considering who made it. But don’t be offended,” he added, “I shall
prize it all the more for its being made by a gentleman, and no
blacksmith; and so will my governor, when I show it to him. I shan’ t
let it remain where it is, but will keep it, as a remembrance of you, as
long as I live.” He then again rubbed his hands with great glee, and
said: “I will now go and see after my horses, and then to breakfast,
partner, if you please”. Suddenly, however, looking at his hands, he
said, “Before sitting down to breakfast I am in the habit of washing
my hands and face; I suppose you could not furnish me with a little
soap and water”. “As much water as you please,” said I, “but if you
want soap, I must go and trouble the young gentlewoman for some.”
“By no means,” said the postillion, “water will do at a pinch.” “Follow
me,” said I, and leading him to the pond of the frogs and newts, I
said, “this is my ewer; you are welcome to part of it—the water is so
soft that it is scarcely necessary to add soap to it;” then lying down
on the bank, I plunged my head into the water, then scrubbed my
hands and face, and afterwards wiped them with some long grass
which grew on the margin of the pond. “B ravo,” said the postillion, “I
see you know how to make a shift”: he then followed my example,
declared he never felt more refreshed in his life, and, giving a bound,
said, “he would go and look after his horses”.
We then went to look after the horses, which we found not much the
worse for having spent the night in the open air. My companion

again inserted their heads in the corn-bags, and, leaving the animals
to discuss their corn, returned with me to the dingle, where we found
the kettle boiling. W e sat down, and Belle made tea, and did the
honours of the meal. The posti llion was in high spirits, ate heartily,
and, to Belle’s evident satisfaction, declared that he had never drank
better tea in his life, or indeed any half so good. B reakfast over, he
said that he must now go and harness his horses, as it was high time
for him to return to his inn. B elle gave him her hand and wished him
farewell. The postillion shook her hand warmly, and was advancing
close up to her—for what purpose I cannot say—whereupon Belle,
withdrawing her hand, drew herself up with an air which caused the
postillion to retreat a step or two with an exceedingly sheepish look.
Recovering himself, however, he made a low bow, and proceeded up
the path. I at tended him, and helped to harness his horses and put
them to the vehicle; he then shook me by the hand, and taking the
reins and whip mounted to his seat; ere he drove away he thus
addressed me: “If ever I forget your kindness and that of the young
woman below, dash my buttons. If ever either of you should enter
my inn you may depend upon a warm welcome, the best that can be
set before you, and no expense to either, for I will give both of you
the best of characters to the governor, who is the very best fellow
upon all the road. As for your linch-pin, I trust it will serve till I get
home, when I will take it out and keep it in remembrance of you all
the days of my life”: then giving the horses a jerk with his reins, he
cracked his whip and drove off.
I returned to the dingle, Belle had removed the breakfast things, and
was busy in her own encampment. Nothing oc curred, worthy of
being related, for two hours, at the end of which time Belle departed
on a short expedition, and I again found myself alone in the dingle.

Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.
More than just a book-buying platform, we strive to be a bridge
connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.
Join us on a journey of knowledge exploration, passion nurturing, and
personal growth every day!
ebookbell.com

Machine Learning In Information And Communication Technology Proceedings Of Icict 2021 Smit Hiren Kumar Deva Sarma

About This Presentation

Slide Content

Tags

Categories

Download

Quick Actions

Statistics

Related Slideshows

Machine Learning In Information And Communication Technology Proceedings Of Icict 2021 Smit Hiren Kumar Deva Sarma

About This Presentation

Slide Content

Slide 1

Slide 2

Slide 3

Slide 5

Slide 6

Slide 7

Slide 8

Slide 9

Slide 10

Slide 11

Slide 12

Slide 13

Slide 14

Slide 15

Slide 16

Slide 17

Slide 18

Slide 19

Slide 20

Slide 21

Slide 22

Slide 23

Slide 24

Slide 25

Slide 26

Slide 27

Slide 28

Slide 29

Slide 30

Slide 31

Slide 32

Slide 33

Slide 34

Slide 35

Slide 36

Slide 37

Slide 38

Slide 39

Slide 40

Slide 41

Slide 42

Slide 43

Slide 44

Slide 45

Slide 46

Slide 47

Slide 48

Slide 49

Slide 50

Slide 51

Slide 52

Slide 53

Slide 54

Slide 55

Slide 59

Slide 60

Slide 61

Slide 62

Slide 63

Slide 64

Slide 65

Slide 66

Slide 67

Slide 68

Slide 69

Slide 70

Slide 71

Slide 72

Slide 73

Slide 74

Slide 75

Slide 76

Slide 77

Slide 78

Tags

Categories

Download