Computational Methods and Data Engineering:

Proceedings of ICMDE 2020, Volume 1 Vijendra

Singh download

https://textbookfull.com/product/computational-methods-and-data-
engineering-proceedings-of-icmde-2020-volume-1-vijendra-singh/

Download more ebook from https://textbookfull.com

We believe these products will be a great fit for you. Click
the link to download now, or visit textbookfull.com
to discover even more!

Computational Methods and Data Engineering: Proceedings

of ICMDE 2020, Volume 2 Vijendra Singh

https://textbookfull.com/product/computational-methods-and-data-
engineering-proceedings-of-icmde-2020-volume-2-vijendra-singh/

Software Engineering Perspectives in Intelligent

Systems: Proceedings of 4th Computational Methods in
Systems and Software 2020, Vol.1 Radek Silhavy

https://textbookfull.com/product/software-engineering-
perspectives-in-intelligent-systems-proceedings-of-4th-
computational-methods-in-systems-and-software-2020-vol-1-radek-
silhavy/

Data Management, Analytics and Innovation: Proceedings

of ICDMAI 2020, Volume 1 Neha Sharma

https://textbookfull.com/product/data-management-analytics-and-
innovation-proceedings-of-icdmai-2020-volume-1-neha-sharma/

CFA 2020 Level 1 Volume 1 Ethics and Quantitative

Methods Various

https://textbookfull.com/product/cfa-2020-level-1-volume-1-
ethics-and-quantitative-methods-various/
Software Engineering Perspectives in Intelligent
Systems: Proceedings of 4th Computational Methods in
Systems and Software 2020, Vol.2 Radek Silhavy

https://textbookfull.com/product/software-engineering-
perspectives-in-intelligent-systems-proceedings-of-4th-
computational-methods-in-systems-and-software-2020-vol-2-radek-
silhavy/

Coulson and Richardson’s Chemical Engineering, Fourth

Edition: Volume 3A: Chemical and Biochemical Reactors
and Reaction Engineering R. Ravi

https://textbookfull.com/product/coulson-and-richardsons-
chemical-engineering-fourth-edition-volume-3a-chemical-and-
biochemical-reactors-and-reaction-engineering-r-ravi/

Data Engineering and Applications Volume 1 Rajesh Kumar

Shukla

https://textbookfull.com/product/data-engineering-and-
applications-volume-1-rajesh-kumar-shukla/

Proceedings of International Conference on

Computational Intelligence and Data Engineering
Proceedings of ICCIDE 2018 Nabendu Chaki

https://textbookfull.com/product/proceedings-of-international-
conference-on-computational-intelligence-and-data-engineering-
proceedings-of-iccide-2018-nabendu-chaki/

Advances in Computational and Bio-Engineering:

Proceeding of the International Conference on
Computational and Bio Engineering, 2019, Volume 1 S.
Jyothi
https://textbookfull.com/product/advances-in-computational-and-
bio-engineering-proceeding-of-the-international-conference-on-
computational-and-bio-engineering-2019-volume-1-s-jyothi/
Advances in Intelligent Systems and Computing 1227

Vijendra Singh
Vijayan K. Asari
Sanjay Kumar
R. B. Patel Editors

Computational
Methods
and Data
Engineering
Proceedings of ICMDE 2020, Volume 1
Advances in Intelligent Systems and Computing

Volume 1227

Series Editor
Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences,
Warsaw, Poland

Advisory Editors
Nikhil R. Pal, Indian Statistical Institute, Kolkata, India
Rafael Bello Perez, Faculty of Mathematics, Physics and Computing,
Universidad Central de Las Villas, Santa Clara, Cuba
Emilio S. Corchado, University of Salamanca, Salamanca, Spain
Hani Hagras, School of Computer Science and Electronic Engineering,
University of Essex, Colchester, UK
László T. Kóczy, Department of Automation, Széchenyi István University,
Gyor, Hungary
Vladik Kreinovich, Department of Computer Science, University of Texas
at El Paso, El Paso, TX, USA
Chin-Teng Lin, Department of Electrical Engineering, National Chiao
Tung University, Hsinchu, Taiwan
Jie Lu, Faculty of Engineering and Information Technology,
University of Technology Sydney, Sydney, NSW, Australia
Patricia Melin, Graduate Program of Computer Science, Tijuana Institute
of Technology, Tijuana, Mexico
Nadia Nedjah, Department of Electronics Engineering, University of Rio de Janeiro,
Rio de Janeiro, Brazil
Ngoc Thanh Nguyen , Faculty of Computer Science and Management,
Wrocław University of Technology, Wrocław, Poland
Jun Wang, Department of Mechanical and Automation Engineering,
The Chinese University of Hong Kong, Shatin, Hong Kong
The series “Advances in Intelligent Systems and Computing” contains publications
on theory, applications, and design methods of Intelligent Systems and Intelligent
Computing. Virtually all disciplines such as engineering, natural sciences, computer
and information science, ICT, economics, business, e-commerce, environment,
healthcare, life science are covered. The list of topics spans all the areas of modern
intelligent systems and computing such as: computational intelligence, soft comput-
ing including neural networks, fuzzy systems, evolutionary computing and the fusion
of these paradigms, social intelligence, ambient intelligence, computational neuro-
science, artificial life, virtual worlds and society, cognitive science and systems,
Perception and Vision, DNA and immune based systems, self-organizing and
adaptive systems, e-Learning and teaching, human-centered and human-centric
computing, recommender systems, intelligent control, robotics and mechatronics
including human-machine teaming, knowledge-based paradigms, learning para-
digms, machine ethics, intelligent data analysis, knowledge management, intelligent
agents, intelligent decision making and support, intelligent network security, trust
management, interactive entertainment, Web intelligence and multimedia.
The publications within “Advances in Intelligent Systems and Computing” are
primarily proceedings of important conferences, symposia and congresses. They
cover significant recent developments in the field, both of a foundational and
applicable character. An important characteristic feature of the series is the short
publication time and world-wide distribution. This permits a rapid and broad
dissemination of research results.
** Indexing: The books of this series are submitted to ISI Proceedings,
EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink **

More information about this series at http://www.springer.com/series/11156

Vijendra Singh Vijayan K. Asari
• •

Sanjay Kumar R. B. Patel

•

Editors

Computational Methods
and Data Engineering
Proceedings of ICMDE 2020, Volume 1

123
Editors
Vijendra Singh Vijayan K. Asari
School of Computer Science Department of Electrical and Computer
University of Petroleum and Energy Studies Engineering
Dehradun, Uttarakhand, India University of Dayton
Dayton, OH, USA
Sanjay Kumar
Department of Computer Science R. B. Patel
and Engineering Department of Computer Science
SRM University Delhi-NCR and Engineering
Sonepat, Haryana, India Chandigarh College of Engineering
and Technology
Chandigarh, Punjab, India

ISSN 2194-5357 ISSN 2194-5365 (electronic)

Advances in Intelligent Systems and Computing
ISBN 978-981-15-6875-6 ISBN 978-981-15-6876-3 (eBook)
https://doi.org/10.1007/978-981-15-6876-3
© Springer Nature Singapore Pte Ltd. 2021
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

We are pleased to present Springer Book entitled Computational Methods and Data
Engineering, which consists of the Proceedings of the International Conference on
Computational Methods and Data Engineering (ICMDE 2020), Volume 1 papers.
The main aim of the International Conference on Computational Methods and
Data Engineering (ICMDE 2020) was to provide a platform for researchers and
academia in the area of computational methods and data engineering to exchange
research ideas, results and collaborate together. The conference was held at the SRM
University, Sonepat, Haryana, Delhi-NCR, India, from January 30 to 31, 2020.
All the 49 published chapters in the Computational Methods and Data Engineer-
ing book have been peer reviewed by three reviewers drawn from the scientific
committee, external reviewers and editorial board depending on the subject matter
of the chapter. After the rigorous peer-review process, the submitted papers were
selected based on originality, significance and clarity and published as chapters.
We would like to express our gratitude to the management, faculty members and
other staff of the SRM University, Sonepat, for their kind support during the
organization of this event. We would like to thank all the authors, presenters and
delegates for their valuable contribution in making this an extraordinary event.
We would like to acknowledge all the members of honorary advisory chairs,
international/national advisory committee members, general chairs, program chairs,
organization committee members, keynote speakers, the members of the technical
committees and reviewers for their work.
Finally, we thank series editors, Advances in Intelligent Systems and
Computing, Aninda Bose and Radhakrishnan for their high support and help.

Editors
Dehradun, India Vijendra Singh
Dayton, USA Vijayan K. Asari
Sonepat, India Sanjay Kumar
Chandigarh, India R. B. Patel

v
Contents

Content Recommendation Based on Topic Modeling . . . . . . . . . . . . . . . 1

Sachin Papneja, Kapil Sharma, and Nitesh Khilwani
Hybrid ANFIS-GA and ANFIS-PSO Based Models for Prediction
of Type 2 Diabetes Mellitus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Ratna Patil, Sharvari Tamane, and Nirmal Rawandale
Social Network Analysis of YouTube: A Case Study on Content
Diversity and Genre Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Shubham Garg, Saurabh, and Manvi Breja
Feature Extraction Technique for Vision-Based Indian Sign Language
Recognition System: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Akansha Tyagi and Sandhya Bansal
Feature-Based Supervised Classifier to Detect Rumor
in Social Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Anamika Joshi and D. S. Bhilare
K-harmonic Mean-Based Approach for Testing
the Aspect-Oriented Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Richa Vats and Arvind Kumar
An Overview of Use of Artificial Neural Network in Sustainable
Transport System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Mohit Nandal, Navdeep Mor, and Hemant Sood
Different Techniques of Image Inpainting . . . . . . . . . . . . . . . . . . . . . . . 93
Megha Gupta and R. Rama Kishore
Web-Based Classification for Safer Browsing . . . . . . . . . . . . . . . . . . . . . 105
Manika Bhardwaj, Shivani Goel, and Pankaj Sharma

vii
viii Contents

A Review on Cyber Security in Metering Infrastructure

of Smart Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Anita Philips, J. Jayakumar, and M. Lydia
On Roman Domination of Graphs Using a Genetic Algorithm . . . . . . . 133
Aditi Khandelwal, Kamal Srivastava, and Gur Saran
General Variable Neighborhood Search for the Minimum Stretch
Spanning Tree Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Yogita Singh Kardam and Kamal Srivastava
Tabu-Embedded Simulated Annealing Algorithm for Profile
Minimization Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Yogita Singh Kardam and Kamal Srivastava
Deep Learning-Based Asset Prognostics . . . . . . . . . . . . . . . . . . . . . . . . . 181
Soham Mehta, Anurag Singh Rajput, and Yugalkishore Mohata
Evaluation of Two Feature Extraction Techniques for Age-Invariant
Face Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Ashutosh Dhamija and R. B. Dubey
XGBoost: 2D-Object Recognition Using Shape Descriptors
and Extreme Gradient Boosting Classifier . . . . . . . . . . . . . . . . . . . . . . . 207
Monika, Munish Kumar, and Manish Kumar
Comparison of Principle Component Analysis and Stacked
Autoencoder on NSL-KDD Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Kuldeep Singh, Lakhwinder Kaur, and Raman Maini
Maintainability Configuration for Component-Based Systems
Using Fuzzy Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Kiran Narang, Puneet Goswami, and K. Ram Kumar
Development of Petri Net-Based Design Model for Energy Efficiency
in Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Sonal Dahiya, Ved Prakash, Sunita Kumawat, and Priti Singh
Lifting Wavelet and Discrete Cosine Transform-Based
Super-Resolution for Satellite Image Fusion . . . . . . . . . . . . . . . . . . . . . . 273
Anju Asokan and J. Anitha
Biologically Inspired Intelligent Machine and Its Correlation
to Free Will . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Munesh Singh Chauhan
Weather Status Prediction of Dhaka City Using Machine
Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Sadia Jamal, Tanvir Hossen Bappy, Roushanara Pervin,
and AKM Shahariar Azad Rabby
Contents ix

Image Processing: What, How and Future . . . . . . . . . . . . . . . . . . . . . . . 305

Mansi Lather and Parvinder Singh
A Study of Efficient Methods for Selecting Quasi-identifier
for Privacy-Preserving Data Mining . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Rigzin Angmo, Veenu Mangat, and Naveen Aggarwal
Day-Ahead Wind Power Forecasting Using Machine Learning
Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
R. Akash, A. G. Rangaraj, R. Meenal, and M. Lydia
Query Relational Databases in Punjabi Language . . . . . . . . . . . . . . . . . 343
Harjit Singh and Ashish Oberoi
Machine Learning Algorithms for Big Data Analytics . . . . . . . . . . . . . . 359
Kumar Rahul, Rohitash Kumar Banyal, Puneet Goswami,
and Vijay Kumar
Fault Classification Using Support Vectors for Unmanned
Helicopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Rupam Singh and Bharat Bhushan
EEG Signal Analysis and Emotion Classification
Using Bispectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Nelson M. Wasekar, Chandrkant J. Gaikwad, and Manoj M. Dongre
Slack Feedback Analyzer (SFbA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
Ramchandra Bobhate and Jyoti Malhotra
A Review of Tools and Techniques for Preprocessing
of Textual Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Abhinav Kathuria, Anu Gupta, and R. K. Singla
A U-Shaped Printed UWB Antenna with Three Band Rejection . . . . . . 423
Deepak Kumar, Preeti Rani, Tejbir Singh, and Vishant Gahlaut
Prediction Model for Breast Cancer Detection Using Machine
Learning Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
Nishita Sinha, Puneet Sharma, and Deepak Arora
Identification of Shoplifting Theft Activity Through Contour
Displacement Using OpenCV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
Kartikeya Singh, Deepak Arora, and Puneet Sharma
Proof of Policy (PoP): A New Attribute-Based Blockchain
Consensus Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
R. Mythili and Revathi Venkataraman
x Contents

Real-Time Stabilization Control of Helicopter Prototype by IO-IPD

and L-PID Controllers Tuned Using Gray Wolf Optimization
Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
Hem Prabha, Ayush, Rajul Kumar, and Ankit Lal Meena
Factors of Staff Turnover in Textile Businesses in Colombia . . . . . . . . . 479
Erick Orozco-Acosta, Milton De la Hoz-Toscano, Luis Ortiz-Ospino,
Gustavo Gatica, Ximena Vargas, Jairo R. Coronado-Hernández,
and Jesus Silva
CTR Prediction of Internet Ads Using Artificial Organic Networks . . . . 489
Jesus Silva, Noel Varela, Danelys Cabrera,
and Omar Bonerge Pineda Lezama
Web Platform for the Identification and Analysis of Events
on Twitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Amelec Viloria, Noel Varela, Jesus Vargas,
and Omar Bonerge Pineda Lezama
Method for the Recovery of Indexed Images in Databases
from Visual Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Amelec Viloria, Noel Varela, Jesus Vargas,
and Omar Bonerge Pineda Lezama
Model for Predicting Academic Performance Through Artificial
Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
Jesus Silva, Ligia Romero, Darwin Solano, Claudia Fernandez,
Omar Bonerge Pineda Lezama, and Karina Rojas
Feature-Based Sentiment Analysis and Classification
Using Bagging Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
Yash Ojha, Deepak Arora, Puneet Sharma, and Anil Kumar Tiwari
A Novel Image Encryption Method Based on LSB Technique
and AES Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
Paras Chaudhary
Implementing Ciphertext Policy Encryption in Cloud Platform
for Patients’ Health Information Based on the Attributes . . . . . . . . . . . 547
S. Boopalan, K. Ramkumar, N. Ananthi, Puneet Goswami,
and Suman Madan
Improper Passing and Lane-Change Related Crashes: Pattern
Recognition Using Association Rules Negative Binomial Mining . . . . . . 561
Subasish Das, Sudipa Chatterjee, and Sudeshna Mitra
Sleep Stage and Heat Stress Classification of Rodents Undergoing
High Environmental Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
Prabhat Kumar Upadhyay and Chetna Nagpal
Contents xi

Development of a Mathematical Model for Solar Power Estimation

Using Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
Arjun Viswanath, Karthik Krishna, T. Chandrika, Vavilala Purushotham,
and Priya Harikumar
Cloud Based Interoperability in Healthcare . . . . . . . . . . . . . . . . . . . . . . 599
Rakshit Joshi, Saksham Negi, and Shelly Sachdeva
Non-attendance of Lectures; Perceptions of Tertiary Students:
A Study of Selected Tertiary Institutions in Ghana . . . . . . . . . . . . . . . . 613
John Kani Amoako and Yogesh Kumar Sharma

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623

About the Editors

Dr. Vijendra Singh is working as Professor in the School of Computer Science at

The University of Petroleum and Energy Studies, Dehradun, Uttarakhand, India.
Prior to joining the UPES, he worked with the NCU, Delhi-NCR, India, Mody
University, Lakshmangarh, India, and Asian CERC Information Technology Ltd.
Dr. Singh received his Ph.D. degree in Engineering and M.Tech. degree in
Computer Science and Engineering from Birla Institute of Technology, Mesra,
India. He has 20 years of experience in research and teaching including IT industry.
Dr. Singh major research concentration has been in the areas of data mining, pattern
recognition, image processing, big data, machine learning, and soft computation.
He has published more than 65 scientific papers in this domain. He has served as
Editor-in-Chief, Special Issue, Procedia Computer Science, Vol 167, 2020,
Elsevier; Editor-in-Chief, Special Issue, Procedia Computer Science, Vol 132,
2018, Elsevier; Associate Editor, International Journal of Healthcare Information
Systems and Informatics, IGI Global, USA; Guest Editor, Intelligent Data Mining
and Machine Learning, International Journal of Healthcare Information Systems
and Informatics, IGI Global, USA; Editor-in-Chief, International Journal of Social
Computing and Cyber-Physical Systems, Inderscience, UK; Editorial Board
Member, International Journal of Multivariate Data Analysis, Inderscience, UK;
Editorial Board Member, International Journal of Information and Decision
Sciences, Inderscience, UK.

Dr. Vijayan K. Asari is a Professor in Electrical and Computer Engineering and

Ohio Research Scholars Endowed Chair in Wide Area Surveillance at the
University of Dayton, Dayton, Ohio. He is the Director of the University of Dayton
Vision Lab (Center of Excellence for Computer Vision and Wide Area Surveillance
Research). Dr. Asari had been a Professor in Electrical and Computer Engineering
at Old Dominion University, Norfolk, Virginia, till January 2010. He was the
Founding Director of the Computational Intelligence and Machine Vision
Laboratory (ODU Vision Lab) at ODU. Dr. Asari received the bachelor’s degree in
Electronics and Communication Engineering from the University of Kerala
(College of Engineering, Trivandrum), India, in 1978, the M.Tech. and Ph.D.

xiii
xiv About the Editors

degrees in Electrical Engineering from the Indian Institute of Technology, Madras,

in 1984 and 1994, respectively. Dr. Asari received several teachings, research,
advising, and technical leadership awards. Dr. Asari received the Outstanding
Teacher Award from the Department of Electrical and Computer Engineering in
April 2002 and the Excellence in Teaching Award from the Frank Batten College of
Engineering and Technology in April 2004. Dr. Asari has published more than 480
research papers including 80 peer-reviewed journal papers co-authoring with his
graduate students and colleagues in the areas of image processing, computer vision,
pattern recognition, machine learning, and high-performance digital system archi-
tecture design. Dr. Asari has been a Senior Member of the IEEE since 2001 and is a
Senior Member of the Society of Photo-Optical Instrumentation Engineers (SPIE).
He is a Member of the IEEE Computational Intelligence Society (CIS), IEEE CIS
Intelligent Systems Applications Technical Committee, IEEE Computer Society,
IEEE Circuits and Systems Society, Association for Computing Machinery (ACM),
and American Society for Engineering Education (ASEE).

Dr. Sanjay Kumar is working as Professor in the Computer Science and

Engineering Department, SRM University, India. He received his Ph.D. degree in
Computer Science and Engineering from Deenbandhu Chhotu Ram University of
Science and Technology (DCRUST), Murthal (Sonipat), in 2014. He obtained his
B.Tech. and M.Tech. degrees in Computer Science and Engineering in 1999 and
2005, respectively. He has more than 16 years of academic and administrative
experience. He has published more than 15 papers in the international and national
journals of repute. He has also presented more than 12 papers in the international
and national conferences. His current research area is wireless sensor networks,
machine learning, IoT, cloud computing, mobile computing and cyber, and network
security. He chaired the sessions in many international conferences like IEEE,
Springer, and Taylor & Francis. He is the Life Member of Computer Society of
India and Indian Society for Technical Education.

Prof. R. B. Patel is working as Professor in the Department of Computer Science

and Engineering, Chandigarh College of Engineering and Technology (CCET),
Chandigarh, India. Prior to joining the CCET, he worked as Professor at NIT,
Uttarakhand, India, and Dean, Faculty of Information Technology and Computer
Science, Deenbandhu Chhotu Ram University of Science and Technology, Murthal,
India. His research areas include mobile and distributed computing, machine and
deep learning, and wireless sensor networks. Prof. Patel has published more than
150 papers in international journals and conference proceedings. He has supervised
16 Ph.D. scholars and currently 02 are in progress.
Content Recommendation Based
on Topic Modeling

Sachin Papneja, Kapil Sharma, and Nitesh Khilwani

Abstract With the proliferation in Internet usage and communicating devices,

plenty amount of information is available at user disposal but on other side, it leads
to a challenge to provide the fruitful information to end users. To overcome this
problem, recommendation system plays a decisive role in providing pragmatic infor-
mation to end users at appropriate time. This paper proposes a topic modeling based
recommendation system to provide contents related to end users interest. Recom-
mendation systems are based on different filtering mechanisms which are classified
as content based, collaborative based, knowledge based, utility based and hybrid
filtering, etc. The objective of this research is thus to proffer a recommendation
system based on topic modeling. Benefit of latent Dirichlet allocation (LDA) is to
uncover latent semantic structure from the text documents. By analyzing the contents
using topic modeling, system can recommend the right articles to end users based
on user interest.

Keywords Recommendation system · LDA · Topic modeling · Content filtering ·

Collaborative filtering

1 Introduction

In last few years, with the telecom revolution, Internet has become a powerful tool
which has changed the way user communicate among themselves as well use it in
the professional business. As per year 2018 statistics, there are now more than 4

S. Papneja (B) · K. Sharma

Department of Computer Science & Engineering, Delhi Technological University,
New Delhi, India
e-mail: [email protected]
K. Sharma
e-mail: [email protected]
N. Khilwani
RoundGlass, New Delhi, India
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2021 1

V. Singh et al. (eds.), Computational Methods and Data Engineering,
Advances in Intelligent Systems and Computing 1227,
https://doi.org/10.1007/978-981-15-6876-3_1
2 S. Papneja et al.

billion people around the world using the Internet whereas around 7.5 billion mobile
connections across the globe. As per the assessment, there are close to 1.5 billion
Internet sites on the cyberspace today. Out of the total available sites, less than 200
million are operating. As the number of communicating devices increase rapidly, it
results in the infinite amount of data in the embodiment of text, images, and videos.
Fundamental test is to give the exact data to the user dependent on user intrigue.
Amidst the appearance of Internet network accessibility, user’s propensities for
understanding news or most recent data have alternated from magazine or booklet
to advanced substance. Because of the immense amount of information accessible
on the cyberspace, it is extremely awkward for the end user to have the data acces-
sible according to his/her advantage. Recommender Systems aid conquers this issue
and gives important data or administrations to the user. Various sorts of suggestion
frameworks exist, for example, content based [17], collaborative [13], hybrid [7],
utility based, multi-criteria, context-aware, risk-aware based, each having with their
impediments. Analysts need to utilize distinctive suggestion frameworks dependent
on their exploration territories.
Content-based frameworks attempt to prescribe things like those a given user
has enjoyed before. For sure, the essential procedure performed by a content-based
recommender comprises in coordinating up the characteristics of a client profile in
which inclinations and interests are put away, with the properties of a substance
object (thing), so as to prescribe to the client new intriguing things. Content-based
recommenders exploit solely ratings. Content-based recommenders are capable of
recommending items not yet rated by any user provided by the active user to build
her own profile.
Numerous customary news recommender frameworks utilize collective sifting
to make suggestions dependent on the conduct of clients in the framework. In this
methodology, the presentation of new clients or new things can cause the cold start
issue, as there will be lacking information on these new sections for the communi-
tarian separating to draw any deductions for new clients or things. Content-based
news recommender frameworks developed to address the cold start issue. In any
case, many substance-based news recommender frameworks consider records as a
sack of-words disregarding the shrouded subjects of the news stories.
Individuals have consistently been standing up to with a developing measure of
information, which thusly requests more on their capacities to channel the substance
as indicated by their inclinations. Among the undeniably overpowering measures of
website pages, records, pictures, or recordings, it is never again natural to and what
we truly need. Besides, copy or a few data sources are discovered covering similar
themes. The clients are touchy to the recentness of data and their inclinations are
additionally changing after some time alongside the substance of the Web.
During the previous two decades, the ideas of recommender frameworks have
risen to cure the circumstance. The quintessence of recommender frameworks are
profoundly connected with the broad work in psychological science, guess hypoth-
esis, data recovery, determining speculations, and the board science. The content-
based methodology of suggestion has its foundations in data recovery [18], and
data separating [13] research. Content-based frameworks are planned for the most
Content Recommendation Based on Topic Modeling 3

part to suggest content-based things; the substance in these frameworks is generally

portrayed with keyword. Customized recommender frameworks intend to prescribe
applicable things to users dependent on their watched conduct, e.g., search person-
alization [3], Google News personalization [4], and Yahoo! conduct focusing on [5]
among others.
As of late, topic modeling approach, for example, latent Dirichlet allocation
(LDA) and probabilistic latent semantic analysis (pLSA) helps in examining content
substance includes by uncovering latent topics of each archive from the document
archive. Reason for LDA is to reveal the semantic structure covered up in the docu-
ments which incorporates the word appropriations over the latent subjects and the
inactive point disposal over archives [8]. The principle challenge is the way to suggest
explicit articles from an immense measure of recently accessible data to end customer,
where the chosen commodity should coordinate the buyer intrigue. In this research
work, concocted a recommendation framework dependent on LDA topic modeling.
In the recommended framework, LDA topic modeling is used to uncover topic from
the document related to user hobbies. Once system know the user interest, based on
the information system can recommend the articles related to interest by filtering
required information. One of the significant qualities of probabilistic topic modeling
is the capacity to uncover shrouded relations through the examination of co-event
designs on dyadic perceptions, for example, document-term pairs.

2 Related Researches

The main purpose of Recommender System is to assist users to make accurate deci-
sions without spending too much on searching this vast amount of information.
Traditional Recommender System is designed to recommend meaningful items to
their users. Those items depend on the purpose of the RS, for example, Google
recommends news to people while Facebook recommends people (friends) to people.
Recommender Systems are a sub-class of information retrieval systems and designed
to predict users’ future preferences by analyzing their past interaction with the system.
Usage of Recommender System became more common in recent years.
From the last two decades, Recommender Systems have become the topic of
interest for both academician and for the industry due to increase in overloaded infor-
mation and to provide relevant information to end users [1] by filtering out the infor-
mation. A knowledge-based filtering framework is a data framework intended for
unstructured or semi-organized information [5]. Recommender System may antici-
pate whether a end user would be keen on purchasing a specific item or not. Social
recommendation strategies gather assessment of commodity from numerous people,
and use nearest neighbor procedures to make proposals to a user concerning new
stock [4]. Recommendation system has been largely used in approximation theory
[14], cognitive science [16], forecasting theory, management science. In addition
to Recommender Systems works on the absolute values of ratings, [9] worked on
preference-based filtering, i.e., anticipating the general inclinations of end user.
4 S. Papneja et al.

Xia et al. [19] Suggested content-based recommender framework for E-Commerce

Platform and took a shot at streamlines the coupon picking process and customizes
the suggestion to improve the active clicking factor and, eventually, the conversion
rates. Deng et al. [10] proposed the amalgamation of item rating data that user has
given plus consolidated features of item to propose a novel recommendation model.
Bozanta and Kutlu [6] proposed to gathered client visit chronicles, scene related
data (separation, class, notoriety and cost) and relevant data (climate, season, date
and time of visits) identified with singular client visits from different sources as each
current scene suggestion framework calculation has its own disadvantages. Another
issue is that basic data about setting is not ordinarily utilized in scene suggestion
frameworks. Badriyah et al. [3] utilize proposed framework which suggest property-
related data based on the user action via looking through publicizing content recently
looked by the user. Topic modeling is based on the experience that document consist
of topics whereas topics are congregation of words. Goal of the Topic modeling is
to understand the documents by uncovering hidden latent variables which are used
to describe the document semantic. Latent Semantic Analysis is based on singular
value decomposition (SVD) whereas pLSA is based on probability distribution. LDA
is a Bayesian version of pLSA which uses Dirichlet priors for the document-topic
and word-topic distributions for better generalization.
Luostarinen and Kohonen [12] Studied and compared LDA with other standard
methods such as Naïve Bayes, K-nearest neighbor, regression and regular linear
regression and found out that LDA gives significant improvement in cold start
simulation.
Apaza et al. [2] use LDA by inferring topics from content given in a college course
syllabus for course recommendation to college students from sites such as Coursera,
Udacity, Edx, etc. Pyo et al. [15] proposed unified topic model for User Grouping and
TV program recommendation by employing two latent Dirichlet allocation (LDA)
models. One model is applied on TV users and the other on the viewed TV programs.

3 Background

3.1 Content-Based Recommender Systems

Content-Based (CB) Recommender Systems prescribe things to a user as indicated

by the substance of user’s past inclinations. As such, framework produces proposals
dependent on thing highlights that match with the user profile. The fundamental
procedure can be clarified in two primary advances:
1. Framework makes user profile utilizing user past conduct, all the more exactly
utilizing thing highlights that has been obtained or loved in the past by the user.
2. At that point, framework creates suggestion by breaking down the qualities of
these things and contrasting them and the user profile.
Content Recommendation Based on Topic Modeling 5

Content-based calculation can be comprehended from its name that this strategy for
the most part thinks about thing’s substance. Content-based strategy can be effectively
utilized in thing proposal; however, it necessitates that the applicable traits of the
things can be separated, or at the end of the day it depends on the thing’s substance.
For instance, on the off chance that framework prescribes archives to its users, at that
point the content-based calculation examines reports’ words (content). Be that as it
may, a few things’ highlights cannot be removed effectively, for example, motion
pictures and music, or they can be covered because of security issues consequently
materialness of these techniques is constrained relying upon the idea of the things.
Probabilistic topic models are a suite of methods whose objective is to detect the
concealed topical structure in enormous chronicles of documents.

3.2 Recommender Systems Major Challenges

There are numerous difficulties that recommender framework researchers face today
and those difficulties can influence the algorithm outcome. Some of the challenges
are as follows:
• Data sparsity: Nowadays a great many things are accessible particularly in
online business sites and every day this number is expanding. Along these lines,
finding comparative user (that purchased comparative things) is getting more
enthusiastically. A large portion of the Recommender System calculations are
utilizing user/things closeness to create recommenders. Along these lines, due to
information sparsity calculations may not perform precisely.
• Scalability: Especially, enormous sites have a large number of user and a great
many information. In this way, when planning a Recommender System it ought
to likewise think about the computational expense.
• Cold Start: When new user or information enter the system, system cannot draw
any data hence it cannot produce proposals either. One of the most guileless
answers for the cold start issue is prescribing well known or stylish things to new
users. For instance, in YouTube, when a user has no past video history story it
will prescribe the most famous recordings to this user. In any case, when the user
watches a video then system will have some clue regarding the client’s inclination
and afterward it will prescribe comparative recordings to the past video that the
client has viewed.
• Diversity and accuracy: It is typically viable to prescribe famous things to users.
In any case, users can likewise discover those things independent from anyone else
without a recommender framework. Recommender framework ought to likewise
locate the less famous things however are probably going to be favored by the users
to suggest. One answer for this issue is utilizing mixture suggestion techniques.
• Vulnerability to attacks: Recommender Systems can be focus of a few assaults
attempting to mishandle the Recommender System calculations utilized in the
6 S. Papneja et al.

web-based business sites. Those assaults attempt to trick Recommender System

to wrongly propose foreordained things for benefit.
• The value of time: Customer needs/inclinations will in general change in time.
Be that as it may, most Recommender Systems calculations don’t think about time
as a parameter.
• Evaluation of recommendations: There are a few Recommender System struc-
tured with various purposes and measurements proposed to assess the Recom-
mender System. Notwithstanding, how to pick the one that precisely assesses the
comparing framework is as yet not clear.

3.3 Probabilistic Topic Modeling

Today there are a large amount of articles, site pages, books and web journals acces-
sible on the web. Besides, every day the measure of content reports are expanding
with commitments from informal communities and mechanical improvements. In
this way, finding what we are actually searching for is not a simple assignment as it
used to be and it tends to be very tedious. For instance, for researchers, there are a
million of articles accessible on the web, to locate the related ones is a challenging
task for researchers. It is not practical to peruse every content and compose or classify
them. Along these lines, it is important to utilize programming devices to sort out
them.
For instance, most journals chronicle their issues, putting away every distributed
article, and along these lines, they should store a lot of data. Without utilizing compu-
tational devices arranging such a major unstructured text assortment is unimaginable
by just utilizing human work. In this way, researchers evolve distinctive probabilistic
models for subject revelation from an enormous unstructured text corpus and they
called them probabilistic topic models.
Probabilistic subject models are calculations intended to find the concealed topic
of the article. At the end of the day, they are measurable techniques attempting to
find the shrouded topic of each article by breaking down the recurrence of the words.
The primary thought behind theme models is a presumption that articles are blends
of points (ordinary dispersion) and subjects are typical circulation over words. Topic
models are generative models which fundamentally imply that producing a document
is considered as a probabilistic procedure. This procedure can be clarified in three
fundamental points as pursues:
• Determine an article to be produced.
• Pick topic for every word of the article.
• Draft a word dependent on the topic that has been picked.
Despite the fact that theme models are initially intended to arrange or locate
the shrouded subject of unstructured archives, they have been embraced in a wide
range of spaces with various sorts of information. For instance, they are used in data
retrieval, multimedia retrieval.
Content Recommendation Based on Topic Modeling 7

Probabilistic Topic Modeling comes under the non-supervised learning [11] in

the sense that it does not require antecedent interpretation or document labeling.
In probabilistic modeling, information is exuded from a generative procedure that
incorporates latent variables. This generative procedure characterizes a joint prob-
ability distribution over both the noticed and concealed random variables. It does
not make any earlier supposition how the words are showed up in the document yet
rather what is important to the model is the occurrence of the word is referenced in
the document.

3.4 Latent Dirichlet Allocation

Latent Dirichlet Allocation (LDA) is a three-level hierarchical Bayesian model, in

which every collected item is demonstrated as a limited blend over a basic arrange-
ment of topics and is utilized to reveal topics in a lot of documents. Every topic is, thus,
demonstrated as a limitless blend over a hidden arrangement of topic probabilities.
Document is only having some data about the topic while every topic is portrayed by
dissemination over words. The LDA model is spoken to as a probabilistic graphical
model as shown in Fig. 1.
As it tends to be seen from the diagram that there are three unique degrees of
factors and parameters:
• First level is corpus level parameters and they are examined in the first place for
example before start producing the corpus.
• Second level is record level factors and they are tested once for producing each
archive.
• Third level factors are word-level factors and they are created for each expression
of all records in the corpus.

N
α θ z W
M

Fig. 1 LDA graphical model

8 S. Papneja et al.

In Fig. 1, document is described by M though each document is succession of N

words where word is signified by w and topic variable in document is characterized
by z. The parameters α and β are corpus-level parameters and are inspected once
during the time spent creating a corpus. The factors θ is document level variable,
examined once per document. Lastly, the factors z and ware word-level factors and
are examined once for each word in each document.

4 Proposed System

To provide contents related to user interest, each article related to interest is consid-
ered as document. LDA is used to find out the semantic structure concealed in the
document. LDA provided us a topic distribution for each interest area, so this learning
will help to recommend the related article to end user based on the user interest. LDA
consider each document as collection of topics in a certain distribution and each topic
as a collection of keywords. Once number of topics is decided as input to LDA algo-
rithm, it firstly rearranges the topic proportion within the document and keyword
distribution with in a topic to have a good configuration of topic-keyword.
Accuracy of LDA algorithm depends on some key factors:
1. Quality of input text.
2. Number and variety of topics.
3. Tuning parameters.
In our experiment, we have taken three different topics (cooking, cricket and
bodybuilding) as a user interest for an input to LDA algorithm. Data is gathered
from different websites by writing a crawler in python. Before inputting the data to
the LDA algorithm. All collected data has been cleaned by removing stop words,
removing emails, remove new line characters and remove distracting single quotes.
Once data is preprocessed, now all the sentences are converted into words. To have
more accuracy build the bigram model and performed the lemmatization on the words
followed by removing words whose count is either less than 15% or more than 50%
of the words. Now corpus will be created. Now the preprocessed data is separated
into training set and test set. Once model is prepared with the training set, model
accuracy is checked using the test data.
In Fig. 2, all the three topics are well segregated and have a keywords weight age
for all the three topics.

5 Conclusion and Future Scope

In this paper, content recommendation based on topic modeling is studied and imple-
mented. Implementation is performed on python by considering document related
to three topics and accuracy achieved is 89%. In the future, work will be extended
Content Recommendation Based on Topic Modeling 9

Fig. 2 Topic 1 most relevant terms

by considering more number of different topic documents and system will provide
personalized content to the end users.

References

1. Adomavicius G, Tuzhilin A (2005) Toward the next generation of recommender systems: a

survey of the state-of-the-art and possible extension. Ieee Trans Knowl Data Eng 17(6):734–749
2. Apaza RG, Cervantes EV, Quispe LC, Luna JO (2014) Online courses recommendation based
on LDA. In: Symposium on information management and big data—SIMBig 2014. Peru, p 7
3. Badriyah T, Azvy S, Yuwono W, Syarif I (2018) Recommendation system for property
search using content based filtering method. In: International Conference on Information and
Communications Technology. Yogyakarta
4. Basu C, Hirsh H, Cohen W (1998) Recommendation as classification: using social and content-
based information in recommendation. Am Assoc Artif Intell, p 7. (USA)
5. Belkin NJ, Croft WB (1992). Information filtering and information retrieval: two sides of the
same coin? Commun ACM 35(12):29–38
6. Bozanta A, Kutlu B (2018) HybRecSys: content-based contextual hybrid venue recommender
system. J Inf Sci 45(2)
7. Burke R (2007) Hybrid web recommender systems. Springer-Verlag, Berlin
8. Chang TM, Hsiao W-F (2013) LDA-based personalized document recommendation. Pacific
Asia Conf Inf Sys 13
9. Cohen WW, Schapire RE, Singer Y (1999) Learning to order things. J Artif Intell Res 10:243–
270
10. Deng F, Ren P, Qin Z, Huang G, Qin Z (2018) August). Leveraging image visual features in
content-based recommender system, Hindawi Scientific Programming, p 8
11. Duda RO, Hart PE, Stork DG (2001) Pattern classification. Wiley
12. Luostarinen T, Kohonen O (2013) Using topic models in content-based news recommender
systems. In: 19th Nordic conference of computational linguistics. Oslo, Norway, p 13
10 S. Papneja et al.

13. Pazzani MJ, Billsus D (2007) Content-based recommendation systems. The Adaptive Web.
Berlin, pp 325–341
14. Powell M (1981) Approximation theory and methods. In: Press CU (ed) Press Syndicate of the
University of Cambridge, New York, USA
15. Pyo S, Kim M, Kim E (2014) LDA-based unified topic modeling for similar TV user grouping
and TV program recommendation. IEEE Trans Cybern 16
16. Rich E (1979) User modeling via stereotypes. Elsevier 3(4):329–354
17. Sarkhel JK, Das P (2010) Towards a new generation of reading habits in Internet Era. In: 24th
national seminar of IASLIC. Gorakhpur University, U.P, pp 94–102
18. Su X, Khoshgoftaar TM (2009) A survey of collaborative filtering techniques. Adv Artif Intell
2009(421425), 19
19. Xia Y, Fabbrizio GD, Vaibhav S, Datta A (2017) A content-based recommender system for
e-commerce offers and coupons. SIGIR eCom. Tokyo, p 7
Hybrid ANFIS-GA and ANFIS-PSO
Based Models for Prediction of Type 2
Diabetes Mellitus

Ratna Patil, Sharvari Tamane, and Nirmal Rawandale

Abstract Type- Diabetes Mellitus (T2DM), a major threat to developing as well

as developed countries, can be easily controlled to a large extent through lifestyle
modifications. Diabetes increases the risk of developing various health as well as
financial problems to cure these health complications. The health complications are
stroke, myocardial infarction, and coronary artery disease. Nerve, muscle, kidney
and retinal damage have distressing impact on the life of a diabetic patient. It is
the need of the hour to halt the epidemic of T2DM in the early stage. Data science
approaches have the potential to predict on medical data. Machine learning is an
evolving scientific field in data science where machines learn mechanically and
improve from experience without any explicit program. Our goal was to develop a
system which can improve performance of a classifier for prediction of T2DM. The
purpose of this work is to implement a hybrid model for prediction by integrating the
advantages of artificial neural net (ANN) and fuzzy logic. Genetic algorithm (GA)
and particle swarm optimization (PSO) have been applied to optimize parameters of
developed predicting model. The proposed scheme used a fuzzification matrix. This
matrix is used to relate the input patterns with a degree of membership to different
classes. The specific class is predicted based on the value of degree of membership
of a pattern. We have analyzed the proposed method and previous research in the
literature. High accuracy was achieved using the ANFIS-PSO approach.

Keywords Machine learning · Fuzzy system · Diabetes mellitus · Particle swarm

intelligence approach · Adaptive neuro-fuzzy inference system (ANFIS)

R. Patil (B)
Noida Institute of Engineering and Technology, Greater Noida, Uttar Pradesh, India
e-mail: [email protected]
S. Tamane
Jawaharlal Nehru Engineering College, Aurangabad, India
e-mail: [email protected]
N. Rawandale
Shri Bhausaheb Hire Government Medical College, Dhule, India
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2021 11

V. Singh et al. (eds.), Computational Methods and Data Engineering,
Advances in Intelligent Systems and Computing 1227,
https://doi.org/10.1007/978-981-15-6876-3_2
12 R. Patil et al.

1 Introduction

Diabetes Mellitus is classified into three types. These are namely Type-I (T1DM),
Type-II (T2DM), and Gestational DM (GDM). T2DM appears to be the most
common form of diabetes in India where more than one crore cases are reported
per year. It is developed if insulin is not produced adequately by the pancreas. The
main contributing factors of T2DM include lifestyle, physical inactivity, obesity,
eating habits, and genetics. In T2DM human body does not use insulin properly. We
have considered T2DM for our study. Several classification algorithms are designed
for classifying the patients as diabetic or healthy. ANFIS has its place in the class of
hybrid structure, termed as neuro-fuzzy systems. ANFIS receives the properties of
neural net as well as fuzzy systems [1]. Neural networks can learn effortlessly from
the input provided but it is hard to understand the knowledge assimilated through
neural net [2]. In contrast, fuzzy-based models are understood very straightforwardly.
Fuzzy inference system (FIS) exploits linguistic terms instead of numeric values and
generates rules in the form of if-then structure. Linguistic variables have values in the
form of words in natural language having degrees of membership. Partial member-
ship is allowed in fuzzy sets, which shows that an element exists in more than one
set partially.
The usage of ANFIS makes the creation of the rule base more adaptive to the state
for modeling and controlling of complex and non-linear problems. In this approach,
the rule base is created by exploiting the neural network systems through the back-
propagation process. To boost its performance, the properties of fuzzy logic are
inherited in this model.
In the proposed method, the fusion of ANFIS with metaheuristic approach
has been done. Metaheuristic algorithm follows repetitive process. Metaheuristic
methods control a subordinate heuristic by exploiting and exploring the search space.
These algorithms are stimulated by seeing the phenomena happening in the nature.
This paper is systematized as follows: Related work done by other researchers
is discussed in Sect. 2. Section 3 includes discussion and construction of ANFIS
process. Discussion on GA is represented in Sect. 4 and PSO is depicted in Sect. 5.
Section 6 presents the building of proposed algorithm. Experimental results are
discussed and results obtained are compared in Sect. 7. Lastly, in Sect. 8 concluding
remarks are made.

2 Related Work

ANFIS has been used commonly as an effective tool for prediction due to its learning
abilities and this approach facilitates rapid adaptation to deviations in systems which
directed to robust groundwork for research. In this background work done by other
researchers is presented here.
Hybrid ANFIS-GA and ANFIS-PSO Based Models … 13

Author Soumadip Ghosh have has analyzed the performance of three different
techniques NFS, RBFNN, and ANFIS widely used in Data Mining [3]. Performance
was analyzed based on root mean square error (RMSE), Kappa statistic, F-measure,
accuracy percentage on ten standard datasets from UCI. The results suggest that
ANFIS has RMSE value of 0.4205.
Author Alby in his paper has developed ANFIS with GA and General Regression
Neural Network (GRNN) for prediction of Type-II DM [4]. Using ANFIS with GA
accuracy was 93.49% and accuracy was 85.49% with GRNN classifier.
Authors Ratna, Sharvari Tamne have done the comparison and analysis of logistic
regression (LR), decision tree, K nearest neighbors (KNN), gradient boost, Gaussian
Naïve Bayes, MLP, support vector machine (SVM), and random forest algorithms
[5]. In this study, they have stated the strength and limitations of existing work.
Author Sinan Adnan Diwan Alalwan has carried out a detailed literature survey
on different methods for predicting T2DM [6]. In his work, he has suggested random
forest method and self-organizing map for improving the accuracy of prediction.
Several authors have used PCA technique for dimensionality reduction of dataset.
Authors Ratna et al. have used PCA for dimensionality reduction technique followed
by KMeans in their study and have shown that performance was improved [7, 8].
Author Murat et al. used PCA followed by ANFIS for diagnosing diabetes [9].
Author Quan Zou has implemented three classifiers using random forest, decision
tree, and neural network methods. He has analyzed and compared these classifiers
on PIMA and Luzhou dataset [10]. The study shows that random forests are better
than the other two. For dimensionality reduction PCA and minimum redundancy
maximum relevance (mRMR) were employed. But the result shows that accuracy
was 0.8084 which was better when all the features were used with random forest.
Authors Patil and Tamane have developed the genetic algorithm for feature selec-
tion with K nearest neighbor (KNN) and Naïve Bayes approach [11]. Though both
the models have improved the accuracy of the prediction with reduced feature set,
GA + KNN have got the better results than GA + Naïve Bayes. In GA + KNN
approach, validation accuracy has been improved from 74% to 83%.

3 ANFIS

ANFIS is a fuzzy inference system introduced by Jang, 1993. It is implemented in

the framework of adaptive systems. ANFIS architecture is depicted in Fig. 1. ANFIS
network has two membership functions. Inputs are converted to fuzzy values using
input membership function. Generally used input membership functions are Trian-
gular, Trapezoidal, Gaussian. Fuzzy output of FIS is mapped to crisp value by output
membership functions. Tuning of parameters related with the membership function
is completed during the learning phase. Gradient vector is used for computation of
these parameters and their tuning. For a specified set of parameters, gradient vector
actually computes a measure of how fine the FIS has modeled the provided data.
After getting the gradient vector one of various optimization method can be used
14 R. Patil et al.

Fig. 1 5-layered architecture of ANFIS

for adjusting the parameters for minimizing error measure. This degree of error is
generally calculated by the sum of the squared difference between actual and wanted
outputs.
For approximation of membership function parameters, ANFIS employs either
back-propagation or combination least squares estimation with back-propagation.
Fuzzy rules are created using Sugeno-type fuzzy system on a specified dataset. A
typical form of Sugeno fuzzy rule is:
IF I 1 is Z 1
AND I 2 is Z 2
.....
AND I m is Z m
THEN y = f (I 1 , I 2 ,…, I m )
Where,
I 1 , I 2 ,…, I m are input variables;
Z 1 , Z 2 ,…, Z m are fuzzy sets.
There are five layers with different functions in ANFIS architecture. These
layers are called as fuzzification, product, normalization, de-fuzzy, and output layer
sequentially. Equations (1) to (6) depict function of each layer.
Layer 1: It is a fuzzy layer where the crisp signal is given as input to the ith node.
This node is linked with a linguistic label Ai or else Bi−2 . The function computes
the membership value of the input. The input layer calculates the output from all the
nodes by applying Eqs. (1) and (2).

O1, i = μ Ai (X ), where i = 1, 2 (1)

Hybrid ANFIS-GA and ANFIS-PSO Based Models … 15

O1, i = μ Bi−2 (Y ), where i = 3, 4 (2)

In Eqs. (1) and (2) the inputs to ith node are given by X, Y and Ai , Bi are representing
linguistic symbols. μAi is the membership function of Ai .
Layer 2: All the nodes in this product layer are fixed nodes characterized as . A
rule neuron computes firing strength W i by the product of all the incoming signals
by Eq. (3). Each node output implies the firing strength of a rule.

O2,i = Wi = min {μ Ai (X ), μ Bi (Y )}, where i = 1, 2 (3)

Layer 3: Every node in normalization layer calculates normalized firing strength of

a given rule. It is proportion of the firing strength of specified rule to the summation
of firing strengths of all rules. It indicates the involvement of a given rule to the
ultimate result. Consequently, the output from ith neuron in layer 3 is calculated by
Eq. (4).

Wi
O3,i = Wi = , where i = 1, 2 (4)
(W1 + W2 )

Layer 4: Each neuron in the defuzzification layer computes the weighted

consequential value of a certain rule by Eq. (5).

O4,i = Wi f i = Wi ( pi x + qi y + ri ), where i = 1, 2 (5)


Layer 5: The output layer has a single node. This is a fixed node having label .
It computes the overall ANFIS output by adding the outputs from all the neurons in
the defuzzification layer as in Eq. (6).

O5,1 = Wi f i (6)
i

4 Genetic Algorithm (GA)

They are generally used to produce solutions for optimization and exploration tasks.
GA simulates “survival of-the-fittest” between individuals of succeeding genera-
tion for problem-solving. Genetic algorithms use methods inspired by evolutionary
biology such as selection, inheritance, alteration, and recombination. Pseudocode of
GA is given below:
1. Select initial population.
2. Compute the fitness of every candidate in the populace.
3. Repetition of the next steps (a–e) until termination condition is satisfied.
16 R. Patil et al.

a. High ranking entities are selected for reproduction.

b. Use recombination operator to yield next generation.
c. The resultant offspring is mutated.
d. Evaluate the offspring.
e. Substitution of low ranked chunk of populace with the reproduced descend
ants.

5 Particle Swarm Optimization (PSO)

Kennedy and Eberhart developed PSO in 1995. It is a stochastic optimization method.

The concept of PSO is analogous to flight of birds in hunt of food. It is an evolutionary
optimization method built on the movement and intellect of swarms [12]. PSO is
a population-based searching process where swarm of particles are the searching
agents and position of particle gives solution. Each particle is considered to be a
point (candidate solution) in a N-dimensional space which fine-tunes its “flying”
based on its personal flying experience and the flying experience of other particles.
This concept is represented in Fig. 2.
PSO has found its way in modeling of biological and sociological behavior like
group of birds looking for food cooperatively. The PSO has been also extensively
used in population-based hunt approach. In a search space, the position of particle
is changed repetitively until it reaches to the best solution or until the computational
boundaries are reached.
Pseudocode of PSO is mentioned below:

Fig. 2 PSO concept

 Another Random Document on
Scribd Without Any Related Topics
and foreseeing a whimsical half-hour in watching her pant and
stumble, unassisted, to her point.
Elsbeth was dimly aware of her hostess's attitude, and not a little
nettled by it. She waved away cake and toast with a vague idea of
breaking no bread in the enemy's house, but she was not the
woman to resist tea, though Hecate's self brewed it. Fortified, she
returned the empty cup; readjusted her veil, and opened fire.
"My dear Miss Hartill," she began, a shade too cordially, "I've come
round—I do hope you're not too busy; I know how occupied you
always are."
Clare was not at all busy; entirely at Miss Loveday's service.
"Ah, well, I confess I came round in the hope of finding you alone—
in the hope of a quiet chat——"
Clare was expecting no visitors. But would not Miss Loveday take
another cup of tea?
"Oh no, thank you. Though I enjoyed my cup immensely—delicious
flavour. China, isn't it? Alwynne always quotes your tea. Poor
Alwynne—she can't convert me. I've always drunk the other, you
know. Not but that China tea is to be preferred for those who like it,
of course. An acquired taste, perhaps—at least——" She finished
with an indistinct murmur uncomfortably aware that she had not
been particularly lucid in her compliments to Clare's tea.
Might Clare order a cup of Indian tea to be made for Miss Loveday?
It would be no trouble; her maid drank it, she believed.
"Oh, please don't. I shouldn't dream——You know, I didn't originally
intend to come to tea. But you are so very kind. I am sure you are
wondering what brings me."
Clare disclaimed civilly.
"Well, to tell you the truth—I am afraid you will think me extremely
roundabout, Miss Hartill——"
Clare's mouth twitched.
"But it is not an easy subject to begin. I'm somewhat worried about
Alwynne——"
"Again?" Clare had stiffened, but Elsbeth was too nervous to be
observant.
"Oh, not her health. She is splendidly well again—Dene did
wonders." Clare found Elsbeth's quick little unexplained smile
irritating. "No, this is—well, it certainly has something to do with
Dene, too!"
"Indeed," said Clare.
Elsbeth continued, delicately tactless: she was always at her worst
with her former pupil.
"I daresay you are surprised that I consult you, for we need not
pretend, need we, that we have ever quite agreed over Alwynne?
You, I know, consider me old-fashioned——" She paused a moment
for a disclaimer, but Clare was merely attentive. With a little less
suavity she resumed: "And of course I've always thought that you
——But that, after all, has nothing to do with the matter."
"Nothing whatever," said Clare.
"Exactly. But knowing that you are fond of Alwynne, and realising
your great, your very great, influence with her, I felt—indeed we
both felt—that if you once realised——"
"We?"
"Roger. Mr. Lumsden."
"Oh, the gardener at Dene."
"My cousin, Miss Hartill."
"Oh. Oh, really. But what has he to do with Alwynne?"
"My dear, he wants to marry her. Didn't she tell you?" Elsbeth had
the satisfaction of seeing Clare look startled. "Now I was sure
Alwynne had confided the matter to you. Hasn't she just been here?
That is really why I came. I was so afraid that you, with the best of
motives, of course, might incline her to refuse him. And you know,
Miss Hartill, she mustn't. The very man for Alwynne? He suits her in
every way. Devoted to her, of course, but not in the least weak with
her, and you know I always say that Alwynne needs a firm hand.
And between ourselves, though I am the last person to consider
such a thing, he is an extremely good match. I can't tell you, Miss
Hartill, the joy it was to me, the engagement. I had been anxious—I
quite foresaw that Alwynne would be difficult, though I am
convinced she is attached to him—underneath, you know. So I made
up my mind to come to you. I said to myself: 'I am sure—I am quite
sure—Miss Hartill would not misunderstand the situation. I am quite
sure Miss Hartill would not intend to stand in the child's light. She is
far too fond of Alwynne to allow her personal feelings——' After all,
feminine friendship is all very well, very delightful, of course, and I
am only too sensible of your goodness to Alwynne—and taking her
to Italy too—but when it is a question of Marriage—oh, Miss Hartill,
surely you see what I mean?"
Clare frowned.
"I think so. The gard——This Mr. Lumpkin——"
"Lumsden."
"Of course. I was confusing him——Mr. Lumsden has proposed to
Alwynne. She has refused him, and you now wish for my help in
coercing her into an apparently distasteful engagement?"
"Oh no, Miss Hartill! No question of coercion. I think there is no
possible doubt that she is fond of him, and if it were not for you——
But Alwynne is so quixotic."
Clare lifted her eyebrows, politely blank.
"Oh, Miss Hartill—why beat about the bush? You know your
influence with Alwynne. It is very difficult for me to talk to you.
Please believe that I intend nothing personal—but Alwynne is so
swayed by you, so entirely under your thumb; you know what a
loyal, affectionate child she is, and as far as I can gather from what
Roger let fall—for she is in one of her moods and will not confide in
me—she considers herself bound to you by—by the terms of your
friendship. All she would say to Roger was, 'Clare comes first. Clare
must come first'—which, of course, is perfectly ridiculous."
Clare reddened.
"You mean that I, or you, for that matter, who have known Alwynne
for years, must step aside, must dutifully foster this liking for a
comparative stranger."
Elsbeth smiled.
"Well, naturally. He's a man."
"I am sorry I can't agree. Alwynne is a free agent. If she prefers my
friendship to Mr. Lumsden's adorations——"
"But I've told you already, it's a question of Marriage, Miss Hartill.
Surely you see the difference? How can you weigh the most
intimate, the most ideal friendship against the chance of getting
married?" Elsbeth was wholly in earnest.
Clare mounted her high horse.
"I can—I do. There are better things in life than marriage."
"For the average woman? Do you sincerely say so? The brilliant
woman—the rich woman—I don't count them, and there are other
exceptions, of course; but when her youth is over, what is the
average single woman? A derelict, drifting aimlessly on the high seas
of life. Oh—I'm not very clear; it's easy to make fun of me; but I
know what I mean and so do you. We're not children. We both know
that an unmated woman—she's a failure—she's unfulfilled."
Clare was elaborately bored.
"Really, Miss Loveday, the subject does not interest me."
"It must, for Alwynne's sake. Don't you realise your enormous
responsibility? Don't you realise that when you keep Alwynne
entangled in your apron strings, blind to other interests, when you
cram her with poetry and emotional literature, when you allow her to
attach herself passionately to you, you are feeding, and at the same
time deflecting from its natural channel, the strongest impulse of her
life—of any girl's life? Alwynne needs a good concrete husband to
love, not a fantastic ideal that she calls friendship and clothes in
your face and figure. You are doing her a deep injury, Miss Hartill—
unconsciously, I know, or I should not be here—but doing it, none
the less. If you will consider her happiness——"
Clare broke in angrily—
"I do consider her happiness. Alwynne tells you that I am essential
to her happiness."
"She may believe so. But she's not happy. She has not been happy
for a long time. But she believes herself to be so, I grant you that.
But consider the future. Shall she never break away? Shall she
oscillate indefinitely between you and me, spend her whole youth in
sustaining two old maids? Oh, Miss Hartill, she must have her
chance. We must give her what we've missed ourselves."
Clare appeared to be occupied in stifling a yawn. Her eyes were
danger signals, but Elsbeth was not Alwynne to remark them.
"In one thing, at least, I do thoroughly agree with you. I don't think
there is the faintest likelihood of Alwynne's wishing to marry at all at
present, but I do feel, with you, that it is unfair to expect her to
oscillate, as you rhetorically put it, between two old maids. I agree,
too, that I have responsibilities in connection with her. In fact, I
think she would be happier if she were with me altogether, and I
intend to ask her to come and live here. I shall ask her to-night.
Don't you think she will be pleased?"
Clare's aim was good. Elsbeth clutched at the arms of her chair.
"You wouldn't do such a thing."
Clare laughed shrilly.
"I shall do exactly what your Mr. Lumsden wants to do. I'm not poor.
I can give her a home as well as he, if you are so anxious to get her
off your hands. She seems to be going begging."
Elsbeth rose.
"I'm wasting time. I'll say good-bye, Miss Hartill. I shouldn't have
come. But it was for Alwynne's sake. I hoped to touch you, to
persuade you to forego, for her future's sake, for the sake of her
ultimate happiness, the hold you have on her. I sympathised with
you. I knew it would be a sacrifice. I knew, because I made the
same sacrifice two years ago, when you first began to attract her. I
thought you would develop her. I am not a clever woman, Miss
Hartill, and you are; so I made no stand against you; but it was hard
for me. Alwynne did not make it easier. She was not always kind. But
hearing you to-day, I understand. You made Alwynne suffer more
than I guessed. I don't blame her if sometimes it recoiled on me.
You were always cruel. I remember you. The others were always
snails for you to throw salt upon. I might have known you'd never
change. Do you think I don't know your effect on the children at the
school? Oh, you are a good teacher! You force them successfully;
but all the while you eat up their souls. Sneer if you like! Have you
forgotten Louise? I tell you, it's vampirism. And now you are to take
Alwynne. And when she is squeezed dry and flung aside, who will
the next victim be? And the next, and the next? You grow greedier
as you grow older, I suppose. One day you'll be old. What will you
do when your glamour's gone? I tell you, Clare Hartill, you'll die of
hunger in the end."
The small relentless voice ceased. There was a silence. Clare, who
had remained quiescent for sheer amaze at the attack from so
negligible a quarter, pulled herself together. Rather white, she began
to clap her hands gently, as a critic surprised into applause.
"My dear woman, you're magnificent! Really you are. I never
thought you had it in you. The Law and the Prophets incarnate. How
Alwynne will laugh when I tell her. I wish she'd been here. You
ought to be on the stage, you know, or in the pulpit. Have you quite
finished? Quite? Do unburden yourself completely, you won't be
given another opportunity. You understand that, of course? If
Alwynne wishes to see you, she must make arrangements to do so
elsewhere. That is the one condition I shall make. This is the way
out."
Elsbeth rose. She was furious with herself that her lips must tremble
and her hands shake, as she gathered up scarf and reticule; but she
followed her hostess with sufficient dignity.
Clare flung open the door with a gesture a shade too ample.
Elsbeth laughed tremulously as she passed her and crossed the hall.
"Oh, you are not altered," she said, and bent to fumble at the latch.
"But it doesn't impress me. You've not won yet. You count too much
on Alwynne. And you have still to reckon with Mr. Lumsden."
"And his three acres and a cow!" Clare watched her contemptuously.
It did not seem worth while to keep her dignity with Elsbeth. She felt
that it would be a relief to lose her temper completely, to override
this opponent by sheer, crude invective. She let herself go.
"What a fool you are! Do you flatter yourself that you understand
Alwynne? Go back to your Cœlebs and tell him from Alwynne—I tell
you I speak for Alwynne—that he's wasting his time. Let him take his
goods to another market: Alwynne won't buy. I've other plans for
her—she has other plans for yourself. She doesn't want a husband.
She doesn't want a home. She doesn't want children. She wants me
—and all I stand for. She wants to use her talents—and she shall—
through me. She wants success—she shall have it—through me. She
wants friendship—can't I give it? Affection? Haven't I given it? What
more can she want? A home? I'm well off. A brat to play with? Let
her adopt one, and I'll house it. I'll give her anything she wants.
What more can your man offer? But I won't let her go. I tell you, we
suffice each other. Thank God, there are some women who can do
without marriage—marriage—marriage!"
Elsbeth, as if she heard nothing, tugged at the catch. The door
swung open, and she stepped quietly into the sunny passage. Then
she turned to Clare, a grey, angry shadow in the dusk of the hall.
"Poor Clare!" she said. "Are the grapes very sour?"
She pulled-to the door behind her.

Later in the evening, as she sat, flushed, tremulous, utterly joyful

over Roger's telegram, she considered the manner of her exit and
was shocked at herself.
"I don't know what possessed me," said Elsbeth apologetically. "And
if I had only known. It was unladylike—it was unwomanly—it was
unchristian." She shook her head at her mild self in the glass. "But
she made me so angry! If I'd only known that this was coming!" She
fingered the pink envelope. "She'll think I knew. She'll always think I
knew. And then to say what I did? It was unpardonable.
"But I was right, all the same," cried Elsbeth incorrigibly; "and I
don't care. I'm glad I said it—I'm glad—I'm glad!"
 CHAPTER XLVII
The sun slid over the edge of the sweating earth. Its red-hot plunge
into the sea behind the hills was almost audible. The black cloud,
fuming up from its setting-place, was as the steam of the collision.
In great clots and coils it rolled upwards, spreading as it thinned, till
it was a pall of vapour that sheeted all the lemon-coloured sky.
Suddenly a cold wind sprang up, raced down the silent heavens,
and, by way of Eastern Europe and the North Sea and the straight
Roman road that drives down England, tore along the Utterbridge
byways, and into the open window of Clare Hartill's parlour. A touch
of its cold lips on her hair, and brow, and breast, and it was out
again, driving the dust before it.
Clare shivered. She was very tired of waiting.... It was inexplicable
that Alwynne should be late; but Clare with a half laugh, promised
Alwynne to forego her scolding if she would but come.... The dusk
and the wind and the silence were getting on her nerves.... The tick
of the hall clock, for instance, was aggressive, insistent, maddening
in its precise monotony.... Oh, unbearable! With a gesture that was
hysterical in its abandonment, Clare rose suddenly and flung into the
hall, plucked open the clock door, and removed the pendulum. The
released wire waggled foolishly into silence, like an idiot, tongue a-
loll.
As the quiet hunted Clare into her sitting-room again, a little silver
wire flickered down the sky like a scared snake, and for an instant
she saw herself reflected in a convex mirror, a Clare bleached and
shining and askew, like a St. Michael in a stained-glass window.
Dusk and the thunder followed. The storm was beginning.
Clare moved about restlessly. She disliked storms. Her eyes ached,
and she was cramped with waiting, and Alwynne had not come. She
would, of course.... That woman had detained her, purposely, no
doubt, and now there was the storm to delay her.... But Alwynne
would come.... Clare smiled securely.
Again the lightning whipped across the heavens, and thunder roared
in its wake.
Clare went to the window and watched the sky. The pane of glass
was grateful to her hot forehead. She was too tired, too bruised and
shaken by her own recent anger to arrange her thoughts, to pose for
the moment, even to herself—of all audiences the most critical. The
interview with Elsbeth Loveday rehearsed itself incessantly, pricking,
probing, bludgeoning, in crescendo of intonation, innuendo, open
attack, to the final triumphant insult. Triumphant, because true. The
insult could cut through her defences and strike at her very self,
because it was true. Her pride agonised. She had thought herself
shrouded, invulnerable. And yet Elsbeth, whom of all women she
had reckoned negligible, had guessed, had pitied.... Yet even her
enemy was forgotten, as she sat and shuddered at the wound dealt;
plucked and shrank, and plucked again at the arrow-tip rankling in it
still.
The hours had passed in an evil mazement. But Alwynne was to
come.... She thought of Alwynne with shifting passions of relief and
longing and sheer crude lust for revenge. Alwynne would come....
Alwynne would soothe and comfort, intuitive, never waiting for the
cry for help.
And Alwynne should pay.... Oho! Alwynne should pay Elsbeth's debts
... should wince, and shrink, and whiten. Scientific vivisection of one
nerve. Wait a little, Alwynne!—Ah, Alwynne—the dearest—the
beloved—the light and laughter of one's life.... What fool is
whispering that Clare can hurt her?... Alwynne shall see when she
comes, who loves her.... There shall be a welcome, the royalest
welcome she has ever had.... For what in all the world has Clare but
Alwynne, and having Alwynne, has not Clare the world?
Ah, well.... Perhaps, she had not been always good to Alwynne....
To-day, for instance, she might have been kinder.... But Alwynne
always understood.... That was the comfort of Alwynne, that she
always understood.... Why didn't she come? Wasn't there an echo of
a step far down the street?
When Alwynne came, they would make plans.... It would not be
easy to wean the girl from her aunt, at least while they lived in the
same town, the same country.... But one could travel, could take
Alwynne quite away.... Italy.... Greece.... Egypt.... they would go
round the world together, shake off the school and all it stood for....
In a new world, begin a new life.... Why not? She had money
enough to burn.... It would not be hard to persuade Alwynne,
adventurous, infatuate.... Once gone, Elsbeth might whistle for her
niece.... They would talk it over to-morrow ... to-night ... as soon as
Alwynne came....
Was that thunder or a knocking? Rat-tat! Rat-tat! She had not been
mistaken after all.... Alwynne! Alwynne!
And Clare, with an appearance on her that even Alwynne had never
seen, ran like a child to open the door.
On the threshold stood a messenger boy, proffering a telegram. She
took it.
"Any answer, Miss!" for she had offered to close the door.
"Oh, of course!" She frowned, and pulled open the flimsy sheet.
The boy waited. He peered past her, interested in the odd pictures
on the walls, and the glimpse of a table luxuriously set. The minutes
sped. He had soon seen all he could, and began to fidget.
"Any answer, Miss?" he hinted.
"Oh!" said Clare vaguely. "Answer? No. No answer. No answer at all."
The boy knuckled his forehead and clattered away down the
staircase.
Mechanically Clare shut the door, locked and bolted it and secured it
with the chain. Then she returned to the sitting-room and crossed to
her former station by the open window.
The storm was ending in a downpour of furious tropical rain. It beat
in unheeded upon her thin dress and bare neck and the open
telegram in her hands, as, with lips parted and a faint, puzzled
pucker between her brows, she conned over the message—
I cannot come to-night.—I have gone to Dene. I am going to marry
Roger.
She read it and re-read, twisting it this way and that, for it was
barely visible in the wet dusk. It seemed an eternity before its full
meaning dawned upon her. And yet she had known all there was to
know when she confronted the messenger boy (Oh, Destiny is up to
date) and took her sentence from his grimy hand.
I am going to marry Roger.
"Very well, Alwynne!" Clare flung up her head, up and back. Her face
was drowned in the shadows of the crimson curtain, but her neck
caught the last of the light, shone like old marble. The whole soul of
her showed for an instant in its defiant outline, in the involuntary
pulsation that quivered across its rigidity, in the uncontrollable flutter
beneath the chin.
The thin, capable fingers twisted and clenched over the sodden
paper.
She moved at last, spoke into space. Passion, anger, and the cool
contempt of the school-mistress for a mutinous class, mingled
grotesquely in her voice.
"Very well, Alwynne! Just as you please, of course. There is no more
to be said." She tossed away the little ball of paper as she spoke.
She wandered aimlessly about the room; turned to her book-shelves
after a while, and stood a long time, pulling out volume after
volume, opening each at random, reading a page, closing the book
again, letting it slide from her hand, never troubling to replace it.
She was tired at last and turned to her writing-table.
It was piled high with exercise-books, and she corrected a couple
before she swept them also aside.
The rain had not faltered in its swishing downfall. It beat against the
panes, and on to the sill, and dripped down into a pool beneath the
open window.
"She will have to come back on Monday," said Clare suddenly. "She
can't go off like that. There's the school——" She broke off abruptly,
as a gust of wind soughed by.
I cannot come. I have gone to Dene. I am going to marry Roger.
She could hear Alwynne's voice in it, answering.
"But why?" cried Clare piteously. "Why? What is it? What have I
done?"
"S'hush!" sighed the rain. "S'hush!"
"I loved her," cried Clare. "I loved her. What have I done?"
"S'hush!" sobbed the rain. "S'hush! S'hush!"
She turned to the darkening windows, and started, and shuddered
away again, stricken dumb and shaking. A pool of something red
and wet was spreading over the polished boards, and a thin trickle
was stealing forward to her feet.
Blood?
Fool.... The red of the curtains reflected, tingeing a pool of rain-
water.... Blood, nevertheless.... She had forgotten Louise.
What had Alwynne heard? A garbled version of that last interview?
Fool again—unless the dead can speak.... But Alwynne knew....
Something had been revealed to her, suddenly, during their idle
talk.... But when? But how? She had come as a lover ... she had left
as a stranger ... what in any god's name, had she guessed? Clare's
subconscious memory reproduced for her instantly, with
photographic accuracy, details of the scene that she had not even
known she had observed. Alwynne had changed, in an instant,
between a word and a reply.... What was it that Clare had said—
what trifling, teasing nothing, flung out in pure wantonness? But
Alwynne's face, her dear face, had become, for an instant—Clare
strained to the memory—as the face of Louise.... Louise had looked
at her like that, that other day.... What had they seen then, both of
them? Was she Gorgon to bring that look into their faces? Louise—
yes—she could understand Louise.... She did not care to think about
Louise.... But Alwynne—what had she ever done to Alwynne? At
least Alwynne might tell her what she had done.... She would not
submit to it.... She would not be put aside.... She would at least
have justice....
I am going to marry Roger.
Useless! All useless! The struggle was over before she had known
she was fighting.... She knew that in Alwynne's life there was no
longer any part for her. And Clare had travelled far that evening, to
phrase it thus. Sharing was a strange word for her to use. But she
recognised dully that even sharing was out of her power. What had
she to do with a husband, and housewifery, and the bearing of
children? Alwynne married was Alwynne dead.
Alwynne in love.... Alwynne married.... Alwynne putting any living
thing before Clare! She broke into bitter laughter at the idea. What
had happened? What had Clare done or left undone? She realised
grimly that of this at least she might be sure—it had been her own
doing.... No influence could have wrought against her own....
Alwynne, at least, was where she was, because Clare had sent her,
not because another had beckoned.... And that was the comfort she
had stored up for herself, to last her in the lean years to come....
What was the use of regretting?
Alwynne was gone.... Then forget her.... There were other fish in the
sea.... There was a promising class this term.... That child in the
Fourth.... She wondered if Alwynne had noticed her.... She must ask
Alwynne.... Alwynne had gone away, had gone to Dene, was going
to marry Roger....
Well, there was always work.... Where was that letter to Miss
Marsham?
She moved stiffly in her seat, lit a candle, and drew towards her the
half-written sheet that lay open on the blotter. She re-read it.
You will, I am sure, understand how much I appreciate your offer of
the partnership, but after much consideration I have decided——
She hesitated, crossed out the but and wrote an and above it, and
continued—
—to accept it. I will come to tea to-morrow, as you kindly suggest.
She finished the letter, signed it, stamped and addressed, and sat
idle at last, staring down at it.
The neat handwriting danced, and flickered, and grew dim.
With an awkward gesture she put her hands to her eyes, and
brought them away again, wet. She smiled at that, a twisted,
mocking smile. She supposed she was crying.... She did not
remember ever having done such a thing....
So her future was decided.... It was to be work and loneliness—
loneliness and work ... because, it seemed, she had no friends left....
Yet Alwynne had promised many things.... What had she done to
Alwynne? What had she done?
She turned within herself and reviewed her life as she remembered
it, thought by thought, word by word, action by action. Faces rose
about her, whispering reminders, forgotten faces of the many who
had loved her: from her old nurse, dead long ago, to Louise, and
Alwynne, and foolish Olivia Pring.
The candle at her elbow flared and dribbled, and died at last with a
splutter and a gasp. She paid no heed.
When the dawn came, she was still sitting there, thinking—thinking.
March 1914—September 1915.
THE END
PRINTED IN THE UNITED STATES OF AMERICA
 Transcriber's Note:
Punctuation and formatting markup have been normalized.
Apparent printer's errors have been retained, unless stated below.
Page 22, "critise" changed to "criticise". ("Excuse me, Miss Vigers,
but I hardly see that it is your business to criticise my way of
teaching.")
Page 26, "inacessible" changed to "inaccessible". (Miss Hartill, who
had been, indeed, surrounded, inaccessible, from the instant of
her entrance until the prayer bell rang, did not look her way a
second time.)
Page 29, "Tallyerand" changed to "Talleyrand". (Marengo—
Talleyrand—never heard of 'em!)
Page 30, "returned" changed to "return". (But to return to
Napoleon and the Lower Third——)
Page 31, "warned" changed to "warmed". (And how it warmed the
cockles of one's heart to her!)
Page 43, "all all" changed to "all". (Clare thanked the gods of her
unbelief, and, relaxing all effort, settled herself to enjoy to the full
the cushioning sense of security;)
Page 47, "shouldnt'" changed to "shouldn't". (Well, I thought I
shouldn't get it done under forty—an essay on The Dark Tower.)
Page 83, "scretly" changed to "secretly". (and she would pay any
price for apple-wood, ostensibly for the quality of its flame,
secretly for the mere pleasure of burning fuel with so pleasant a
name;)
Page 88, "a a" changed to "a". (She could not believe in simplicity
combined with brains: a simple soul was necessarily a simpleton in
her eyes.)
Page 89, "negligble" changed to "negligible". (So that negligible
and mouse-like woman had been aware—all along ...)
Page 100, "eucalyplyptus" changed to "eucalyptus". (Before the
evening was over Alwynne reeked of eucalyptus.)
Page 108, "Clarke" changed to "Clare". ("Of course not," said
Clare, with grave sympathy.)
Page 135, "Louise's" changed to "Clare's". (And Alwynne's eyes
grew big, and she forgot all about Louise, as Clare's "loveliest
voice" read out the rhyme of The River.)
Page 152, "Cnythia" changed to "Cynthia". ("And yet it bores her
too——" parenthesised Cynthia shrewdly.)
Page 155, "Wail" changed to "Wait". ("Wait till you get a best
boy.")
Page 186, "then" changed to "them". ("You begin by being
heavenly to people—and then you tantalise them.")
Page 250, "phrase" changed to "phase". (Elsbeth, not unused to
disillusionment and hopes deferred, could sigh and smile and
acquiesce, knowing it for the phase that it was and forgiving
Alwynne in advance.)
Page 370, "so" changed to "to". (She had only to say, quite
quietly, that she must do what she felt to be right....)
Page 413, "Alwyne" changed to "Alwynne". (She thought of
Alwynne with shifting passions of relief and longing and sheer
crude lust for revenge.)
*** END OF THE PROJECT GUTENBERG EBOOK REGIMENT OF
WOMEN ***

Updated editions will replace the previous one—the old editions

will be renamed.

Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States
copyright in these works, so the Foundation (and you!) can copy
and distribute it in the United States without permission and
without paying copyright royalties. Special rules, set forth in the
General Terms of Use part of this license, apply to copying and
distributing Project Gutenberg™ electronic works to protect the
PROJECT GUTENBERG™ concept and trademark. Project
Gutenberg is a registered trademark, and may not be used if
you charge for an eBook, except by following the terms of the
trademark license, including paying royalties for use of the
Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such
as creation of derivative works, reports, performances and
research. Project Gutenberg eBooks may be modified and
printed and given away—you may do practically ANYTHING in
the United States with eBooks not protected by U.S. copyright
law. Redistribution is subject to the trademark license, especially
commercial redistribution.

START: FULL LICENSE

Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com