Download Advances In Machine Learning Deep Learning Based Technologies Selected Papers In Honour Of Professor Nikolaos G Bourbakis Vol 2 Learning And Analytics In Intelligent Systems 23 George A Tsihrintzis online ebook texxtbook full chapter pdf

Advances in Machine Learning Deep
Learning based Technologies Selected

Papers in Honour of Professor Nikolaos
G Bourbakis Vol 2 Learning and
Analytics in Intelligent Systems 23
George A. Tsihrintzis (Editor)
Visit to download the full and correct content document:
https://ebookmeta.com/product/advances-in-machine-learning-deep-learning-based-t
echnologies-selected-papers-in-honour-of-professor-nikolaos-g-bourbakis-vol-2-learni
ng-and-analytics-in-intelligent-systems-23-george-a-tsihrintzis/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Advances in Artificial Intelligence based Technologies

Selected Papers in Honour of Professor Nikolaos G
Bourbakis Vol 1 1st Edition Maria Virvou
https://ebookmeta.com/product/advances-in-artificial-
intelligence-based-technologies-selected-papers-in-honour-of-
professor-nikolaos-g-bourbakis-vol-1-1st-edition-maria-virvou/
Machine Learning Techniques and Analytics for Cloud

Security Advances in Learning Analytics for Intelligent
Cloud IoT Systems 1st Edition Chakraborty
https://ebookmeta.com/product/machine-learning-techniques-and-
analytics-for-cloud-security-advances-in-learning-analytics-for-
intelligent-cloud-iot-systems-1st-edition-chakraborty/
Machine Learning and Deep Learning Techniques in

Wireless and Mobile Networking Systems 1st Edition K.
Suganthi
https://ebookmeta.com/product/machine-learning-and-deep-learning-
techniques-in-wireless-and-mobile-networking-systems-1st-edition-
k-suganthi/
Deep Learning in Cancer Diagnostics: A Feature-based

Transfer Learning Evaluation Mohd Hafiz Arzmi
https://ebookmeta.com/product/deep-learning-in-cancer-
diagnostics-a-feature-based-transfer-learning-evaluation-mohd-
hafiz-arzmi/
Machine Learning and Deep Learning in Neuroimaging Data
Analysis 1st Edition Anitha S. Pillai
https://ebookmeta.com/product/machine-learning-and-deep-learning-
in-neuroimaging-data-analysis-1st-edition-anitha-s-pillai/
Design of Intelligent Applications using Machine

Learning and Deep Learning Techniques 1st Edition
Ramchandra Sharad Mangrulkar (Editor)
https://ebookmeta.com/product/design-of-intelligent-applications-
using-machine-learning-and-deep-learning-techniques-1st-edition-
ramchandra-sharad-mangrulkar-editor/
Advances in Deep Learning, Artificial Intelligence and

Robotics: Proceedings of the 2nd International
Conference on Deep Learning, Artificial ... 2020
(Lecture Notes in Networks and Systems) Luigi Troiano
(Editor)
https://ebookmeta.com/product/advances-in-deep-learning-
artificial-intelligence-and-robotics-proceedings-of-the-2nd-
international-conference-on-deep-learning-
artificial-2020-lecture-notes-in-networks-and-systems-luigi-tro/
Recent Advances in Internet of Things and Machine

Learning: Real-World Applications (Intelligent Systems
Reference Library, 215) Valentina E. Balas (Editor)
https://ebookmeta.com/product/recent-advances-in-internet-of-
things-and-machine-learning-real-world-applications-intelligent-
systems-reference-library-215-valentina-e-balas-editor/
Computer Vision and Machine Learning in Agriculture

(Algorithms for Intelligent Systems) Mohammad Shorif
Uddin (Editor)
https://ebookmeta.com/product/computer-vision-and-machine-
learning-in-agriculture-algorithms-for-intelligent-systems-
mohammad-shorif-uddin-editor/
Volume 23
Learning and Analytics in Intelligent

Systems
Series Editors
George A. Tsihrintzis
University of Piraeus, Piraeus, Greece
Maria Virvou
University of Piraeus, Piraeus, Greece
Lakhmi C. Jain
Faculty of Engineering and Information Technology, Centre for Artificial
Intelligence, University of Technology, Sydney, NSW, Australia;, KES
International, Shoreham-by-Sea, UK; Liverpool Hope University,
Liverpool, UK
The main aim of the series is to make available a publication of books in

hard copy form and soft copy form on all aspects of learning, analytics
and advanced intelligent systems and related technologies. The
mentioned disciplines are strongly related and complement one
another significantly. Thus, the series encourages cross-fertilization
highlighting research and knowledge of common interest. The series
allows a unified/integrated approach to themes and topics in these
scientific disciplines which will result in significant cross-fertilization
and research dissemination. To maximize dissemination of research
results and knowledge in these disciplines, the series publishes edited
books, monographs, handbooks, textbooks and conference proceedings.
More information about this series at http://www.springer.com/
series/16172
Editors
George A. Tsihrintzis, Maria Virvou and Lakhmi C. Jain
Advances in Machine Learning/Deep

Learning-based Technologies
Selected Papers in Honour of Professor Nikolaos G.
Bourbakis – Vol. 2
1st ed. 2022
Editors
Department of Informatics, University of Piraeus, Piraeus, Greece
Maria Virvou
Department of Informatics, University of Piraeus, Piraeus, Greece
Lakhmi C. Jain
KES International, Shoreham-by-Sea, UK
ISSN 2662-3447 e-ISSN 2662-3455

Learning and Analytics in Intelligent Systems
ISBN 978-3-030-76793-8 e-ISBN 978-3-030-76794-5
https://doi.org/10.1007/978-3-030-76794-5
© The Editor(s) (if applicable) and The Author(s), under exclusive

license to Springer Nature Switzerland AG 2022
This work is subject to copyright. All rights are solely and exclusively
licensed by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in
any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks,

service marks, etc. in this publication does not imply, even in the
absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general
use.
The publisher, the authors and the editors are safe to assume that the
advice and information in this book are believed to be true and accurate
at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the
material contained herein or for any errors or omissions that may have
been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer

Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham,
Switzerland
Foreword
Machine Learning can be considered as a part of the Artificial
Intelligence field.
In 1959, Arthur Samuel [1, 2] introduced the term Machine Learning

to refer to research efforts to develop algorithms and procedures
which, when incorporated into machines, would allow them to improve
their performance on specific tasks, i.e., to learn in ways that mimic
human learning [3]. More recent efforts have been inspired by
biological neural structures and have being receiving significant
research attention worldwide. These approaches form a sub-area of
Machine Learning, termed Deep Learning, and include various
computing paradigms of the artificial neural network type, such as
convolutional neural networks, recurrent neural networks, and deep
belief networks [4].
In the six decades since publication of Samuel’s nominal paper,
Machine Learning, in general, and Deep Learning, in particular, have
grown into one of the most active research fields worldwide. These
research efforts have met with success in many technological
application areas and increasingly affect many aspects of everyday life,
the workplace, and human relationships [5–9]. Of course, such a broad
impact also comes with risks and threats in security, privacy, safety,
transparency, business, competition, the job market, fundamental
rights, democracy, or even human existence itself [10–12], which are
hard to ignore and care must be taken to prevent them.
Professor Nikolaos G. Bourbakis stands out as one of the main
contributors to various applications of Machine Learning/Deep
Learning throughout his long and fruitful research career at various
posts. Currently, Nikolaos is a Distinguished Professor of Information &
Technology and the Director of the Center of Assistive Research
Technologies (CART) at Wright State University, Ohio, USA, after
receiving a B.S. degree in Mathematics from the National and
Kapodistrian University of Athens, Greece, a Certificate in Electrical
Engineering from the University of Patras, Greece, and a Ph.D. degree in
Computer Engineering and Informatics (awarded with excellence),
from the Department of Computer Engineering & Informatics,
University of Patras, Greece. His many achievements in Machine
Learning/Deep Learning-based Technologies have been recognized via
many distinctions and awards, including elevation to IEEE Fellow
(1996); IEEE Computer Society Technical Research Achievement
Award; Member of the New York Academy of Sciences; Diploma of
Honor in Artificial Intelligence, School of Engineering, University of
Patras, Greece; ASC Outstanding Scientists & Engineers Research
Award; Dr. F. Russ IEEE Biomedical Engineering Award, Dayton Ohio;
Recognition Award for Outstanding Scholarly Achievements and
Contributions in the field of Computer Science, University of Piraeus,
Greece; IEEE EMBS-GR Award of Achievements; IEEE Computer Society
30 years ICTAI Outstanding Service & Leadership Recognition;
Honorary Doctorate Degree of the University of Piraeus, Greece (2020).
Professors George A. Tsihrintzis, Maria Virvou, and Lakhmi C. Jain
recently undertook a dual task. On one hand they are editing a special
book in Prof. Nikolaos G. Bourbakis’ honor and on the other hand they
are attempting to update the relevant research communities, in
computer science-related disciplines, as well as the general reader from
other disciplines, on the most recent advances in Machine
Learning/Deep Learning-based technological applications. They are
handing to us a book consisting of 11 chapters, each of which has been
written by active and recognized researchers and reports on recent
research and development findings. Overall, the book is well structured
as, besides an editorial note (introductory chapter), it has been further
divided into five parts devoted to Machine Learning/Deep Learning in
Socializing and Entertainment (2 chapters), Machine Learning/Deep
Learning in Education (2 chapters), Machine Learning/Deep Learning in
Security (2 chapters), Machine Learning/Deep Learning in Time Series
Forecasting (2 chapters), and Machine Learning in Video Coding and
Information Extraction (2 chapters).
Even though the area of Machine Learning/Deep Learning-based
Technologies is very broad, the editors have managed to cover it
impressively in terms of both breadth and depth. Undoubtedly, readers
with a background in Artificial Intelligence and Computer Science will
find it helpful in their researches. I am confident that interest will also
be stirred among general readers who are seeking to be versed in
current Machine Learning/Deep Learning-based Technologies.
I, thus, highly recommend this timely book to both the Artificial
Intelligence/Computer Science researchers and the general reader.
Further Reading
1. Arthur Samuel, Some Studies in Machine Learning Using the Game
of Checkers. IBM J. 3(3), 210–229 (1959)
2. https://en.wikipedia.org/wiki/Arthur_Samuel
3. J.E. Ormrod, Human Learning, Pearson, 8th edition, ISBN-13: 978-

0134893662, (2021)
4. M. Nielsen, Neural Nets and Deep Learning (2019), http://

neuralnetworksanddeeplearning.com/
5. K. D. Foote, A Brief History of Machine Learning (2019) (https://

www.dataversity.net/a-brief-history-of-machine-learning/)
6. History of Machine Learning, https://www.doc.ic.ac.uk/~jce317/

history-machine-learning.html
7. Introduction to Neural Nets, https://www.doc.ic.ac.uk/~jce317/

introduction-neural-nets.html
8. B. Grossfeld, Deep learning vs machine learning: a simple way to

understand the difference, https://www.zendesk.com/blog/
machine-learning-and-deep-learning/, published on January 23,
2020, last updated on October 12, 2020
9. G. A. Tsihrintzis and L. C. Jain (Eds.), Machine Learning Paradigms

—Advances in Deep Learning-based Technologies, Vol. 18 in
Learning and Analytics in Intelligent Systems (LAIS), Springer, 2020
10. N. Bourbakis, Artificial Intelligence (AI) and its Impact to

Humanity: Immortality or Last Invention, Invited Keynote Lecture,
University of Piraeus, Greece, Feb. 6, (2020)
11. J. Barrat, Our Final Invention: Artificial Intelligence and the End of
the Human Era, Thomas Dunne Books, October 1, 2013, ISBN-13
978-0312622374
12. Chandrasekar Vuppalapati, Democratization of Artificial

Intelligence for the Future of Humanity, CRC Press, ISBN-13: 978-
0367524128, January 17, (2021)
Michalis Zervakis
Preface
A world-recognized researcher can be honored in a variety of ways,
including elevation of his professional status or various prestigious
awards and distinctions. When, additionally, the same researcher has
served as advisor to generations of undergraduate, graduate, and
doctoral students and as mentor to faculty and colleagues, the task of
appropriately honoring him becomes even harder! Perhaps, the best
way to honor this person is to ask former doctoral students, as
well as colleagues and fellow researchers from around the world,
to include some of their recent research results in one or more
high quality volumes edited in his honor. Such an edition indicates
that other researchers are pursuing and extending further what
they have learned from him in research areas where he made
outstanding contributions.
Professor Nikolaos G. Bourbakis has been serving the fields of
Artificial Intelligence (including Machine Learning/Deep Learning) and
Assistive Technologies from various posts for almost fifty years now. He
received a BS in Mathematics from the National and Kapodistrian
University of Athens, Greece, a Certificate in Electrical Engineering from
the University of Patras, Greece, and a Ph.D. in Computer Engineering
and Informatics (awarded with excellence), from the Department of
Computer Engineering & Informatics, University of Patras, Greece.
Dr. Bourbakis (IEEE Fellow-1996) is currently a Distinguished
Professor of Information & Technology and the Director of the Center of
Assistive Research Technologies (CART) at Wright State University,
Ohio, USA. He is the founder and Editor-in-Chief of the International
Journal on Artificial Intelligence Tools, the International Journal on
Monitoring and Surveillance Technology Research (IGI-Global, Publ.),
and the EAI Transactions on Bioengineering & Bioinformatics. He is
also the Founder and Steering Committee Chair of several International
IEEE Computer Society Conferences (namely, ICTAI, ICBIBE, ICIISA),
Symposia and Workshops. He pursues research in Assistive
Technologies, Applied Artificial Intelligence, Bioengineering,
Information Security, and Parallel/Distributed Processing, which is
funded by USA and European government and industry. He has
published extensively in IEEE and International Journals and he has
graduated, as the main advisor, several dozens of doctoral students. His
research work has been internationally recognized and he has received
several prestigious awards, including: IEEE Computer Society Technical
Research Achievement Award; Member of the New York Academy of
Sciences; Diploma of Honor in AI School of Engineering, University of
Patras, Greece; ASC Outstanding Scientists & Engineers Research
Award; Dr. F. Russ IEEE Biomedical Engineering award, Dayton Ohio;
Most Cited Article in Pattern Recognition Journal; IEEE ICTAI and
ICBIBE best paper Awards; Recognition Award for Outstanding
Scholarly Achievements and Contributions in the field of Computer
Science, University of Piraeus, Greece; IEEE EMBS-GR Award of
Achievements; IEEE Computer Society 30 years ICTAI Outstanding
Service & Leadership Recognition; Honorary Doctorate degree of the
University of Piraeus, Greece.
We have been collaborating with Prof. Nikolaos G. Bourbakis for
very many years. Thus, we proposed and undertook with pleasure the
task of editing a special book in his honor. The response from former
mentees, colleagues, and fellow researchers of his has been great!
Unfortunately, page limitations have forced us to limit the works to be
included in the book and we apologize to those authors whose works
were not included. Despite the decision not to include all proposed
chapters in the book, it became apparent that not only one, but three
volumes of the special book had to be developed, each of which would
focus on different aspects of Dr. Nikolaos G. Bourbakis’s research
activities.
The book at hand constitutes the second volume and is devoted to
Advances in Machine Learning/Deep Learning-based
Technologies. While honoring Professor Nikolaos G. Bourbakis, this
book also serves the purpose of exposing its reader to some of the most
significant advances in Machine Learning/Deep Learning-based
technologies. As such, the book is directed towards professors,
researchers, scientists, engineers, and students in computer science-
related disciplines. It is also directed towards readers who come from
other disciplines and are interested in becoming versed in some of the
most Advances in Machine Learning/Deep Learning-based
Technologies. We hope that all of them will find it useful and inspiring
in their works and researches.
We are grateful to the authors and reviewers for their excellent
contributions and visionary ideas. We are also thankful to Springer for
agreeing to publish this book in its Learning and Analytics in
Intelligent Systems series. Last, but not least, we are grateful to the
Springer staff for their excellent work in producing this book.
Maria Virvou
Lefteri Tsoukalas
Anna Esposito
Lakhmi C. Jain
Piraeus, Greece
Piraeus, Greece
Lafayette, Indiana, USA
Vietri, Italy
Sydney, Australia
Contents
1 Introduction to Advances in Machine Learning/Deep Learning-
Based Technologies
George A. Tsihrintzis, Maria Virvou and Lakhmi C. Jain
1.1 Editorial Note
1.2 Book Summary and Future Volumes
References
Part I Machine Learning/Deep Learning in Socializing and
Entertainment
2 Semi-supervised Feature Selection Method for Fuzzy Clustering
of Emotional States from Social Streams Messages
Ferdinando Di Martino and Sabrina Senatore
2.1 Introduction
2.2 The FS-EFCM Algorithm
2.2.1 EFCM Execution:Main Steps
2.2.2 Initial Parameter Setting
2.3 Experimental Results
2.3.1 Dataset
2.3.2 Feature Selection
2.3.3 FS-EFCM at Work
2.4 Conclusion
References
3 AI in (and for) Games
Kostas Karpouzis and George A. Tsatiris
3.1 Introduction
3.2 Game Content and Databases
3.3 Intelligent Game Content Generation and Selection
3.3.1 Generating Content for a Language Education Game
3.4 Conclusions
References
Part II Machine Learning/Deep Learning in Education
4 Computer-Human Mutual Training in a Virtual Laboratory
Environment
Vasilis Zafeiropoulos and Dimitris Kalles
4.1 Introduction
4.1.1 Purpose and Development of the Virtual Lab
4.1.2 Different Playing Modes
4.1.3 Evaluation
4.2 Background and Related Work
4.3 Architecture of the Virtual Laboratory
4.3.1 Conceptual Design
4.3.2 State-Transition Diagrams
4.3.3 High Level Design
4.3.4 State Machine
4.3.5 Individual Scores
4.3.6 Quantization
4.3.7 Normalization
4.3.8 Composite Evaluation
4.3.9 Success Rate
4.3.10 Weighted Average
4.3.11 Artificial Neural Network
4.3.12 Penalty Points
4.3.13 Aggregate Score
4.4 Machine Learning Algorithms
4.4.1 Genetic Algorithm for the Weighted Average
4.4.2 Training the Artificial Neural Network with Back-
Propagation
4.5 Implementation
4.5.1 Instruction Mode
4.5.2 Evaluation Mode
4.5.3 Computer Training Mode
4.5.4 Training Data Collection Sub-mode
4.5.5 Machine Learning Sub-mode
4.6 Training-Testing Process and Results
4.6.1 Training Data
4.6.2 Training and Testing on Various Data Set Groups
4.6.3 Genetic Algorithm Results
4.6.4 Artificial Neural Network Training Results
4.7 Conclusions
References
5 Exploiting Semi-supervised Learning in the Education Field:A
Critical Survey
Georgios Kostopoulos and Sotiris Kotsiantis
5.1 Introduction
5.2 Semi-supervised Learning
5.3 Literature Review
5.3.1 Performance Prediction
5.3.2 Dropout Prediction
5.3.3 Grade Level Prediction
5.3.4 Grade Point Value Prediction
5.3.5 Other Studies
5.3.6 Discussion
5.4 The Potential of SSL in the Education Field
5.5 Conclusions
References
Part III Machine Learning/Deep Learning in Security
6 Survey of Machine Learning Approaches in Radiation Data
Analytics Pertained to Nuclear Security
Miltiadis Alamaniotis and Alexander Heifetz
6.1 Introduction
6.2 Machine Learning Methodologies in Nuclear Security
6.2.1 Nuclear Signature Identification
6.2.2 Background Radiation Estimation
6.2.3 Radiation Sensor Placement
6.2.4 Source Localization
6.2.5 Anomaly Detection
6.3 Conclusion
References
7 AI for Cybersecurity:ML-Based Techniques for Intrusion
Detection Systems
Dilara Gumusbas and Tulay Yildirim
7.1 Introduction
7.1.1 Why Does AI Pose Great Importance for
Cybersecurity?
7.1.2 Contribution
7.2 ML-Based Models for Cybersecurity
7.2.1 K-Means
7.2.2 Autoencoder (AE)
7.2.3 Generative Adversarial Network (GAN)
7.2.4 Self Organizing Map
7.2.5 K-Nearest Neighbors (k-NN)
7.2.6 Bayesian Network
7.2.7 Decision Tree
7.2.8 Fuzzy Logic (Fuzzy Set Theory)
7.2.9 Multilayer Perceptron (MLP)
7.2.10 Support Vector Machine (SVM)
7.2.11 Ensemble Methods
7.2.12 Evolutionary Algorithms
7.2.13 Convolutional Neural Networks (CNN)
7.2.14 Recurrent Neural Network (RNN)
7.2.15 Long Short Term Memory (LSTM)
7.2.16 Restricted Boltzmann Machine (RBM)
7.2.17 Deep Belief Network (DBN)
7.2.18 Reinforcement Learning (RL)
7.3 Open Topics and Potential Directions
7.3.1 Novel Feature Representations
7.3.2 Unsupervised Learning Based Detection Systems
References
Part IV Machine Learning/Deep Learning in Time Series
Forecasting
8 A Comparison of Contemporary Methods on Univariate Time
Series Forecasting
Aikaterini Karanikola, Charalampos M. Liapis and Sotiris Kotsiantis
8.1 Introduction
8.2 Related Work
8.3 Theoretical Background
8.3.1 ARIMA
8.3.2 Prophet
8.3.3 The Holt-Winters Seasonal Models
8.3.4 N-BEATS:Neural Basis Expansion Analysis
8.3.5 DeepAR
8.3.6 Trigonometric BATS
8.4 Experiments and Results
8.4.1 Datasets
8.4.2 Algorithms
8.4.3 Evaluation
8.4.4 Results
8.5 Conclusions
References
9 Application of Deep Learning in Recurrence Plots for
Multivariate Nonlinear Time Series Forecasting
Sun Arthur A. Ojeda, Elmer C. Peramo and Geoffrey A. Solano
9.1 Introduction
9.2 Related Work
9.2.1 Background on Recurrence Plots
9.2.2 Time Series Imaging and Convolutional Neural
Networks
9.3 Time Series Nonlinearity
9.4 Time Series Imaging
9.4.1 Dimensionality Reduction
9.4.2 Optimal Parameters
9.5 Convolutional Neural Networks
9.6 Model Pipeline and Architecture
9.6.1 Architecture
9.7 Experimental Setup
9.8 Results
9.9 Conclusion
References
Part V Machine Learning in Video Coding and Information
Extraction
10 A Formal and Statistical AI Tool for Complex Human Activity
Recognition
Anargyros Angeleas and Nikolaos Bourbakis
10.1 Introduction
10.2 The Hybrid Framework—Formal Languages
10.3 Formal Tool and Statistical Pipeline Architecture
10.4 DATA Pipeline
10.5 Tools for Implementation
10.6 Experimentation with Datasets to Identify the Ideal Model
10.6.1 KINISIS—Single Human Activity Recognition
Modeling
10.6.2 DRASIS—Change of Human Activity Recognition
Modeling
10.7 Conclusions
References
11 A CU Depth Prediction Model Based on Pre-trained
Convolutional Neural Network for HEVC Intra Encoding
Complexity Reduction
Jiaming Li, Ming Yang, Ying Xie and Zhigang Li
11.1 Introduction
11.2 H.265 High Efficiency Video Coding
11.2.1 Coding Tree Unit Partition
11.2.2 Rate Distortion Optimization
11.2.3 CU Partition and Image Texture Features
11.3 Proposed Methodology
11.3.1 The Hierarchical Classifier
11.3.2 The Methodology of Transfer Learning
11.3.3 Structure of Convolutional Neural Network
11.3.4 Dataset Construction
11.4 Experiments and Results
11.5 Conclusion
References
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
G. A. Tsihrintzis et al. (eds.), Advances in Machine Learning/Deep Learning-based
Technologies, Learning and Analytics in Intelligent Systems 23
https://doi.org/10.1007/978-3-030-76794-5_1
1. Introduction to Advances in Machine

Learning/Deep Learning-Based
Technologies
George A. Tsihrintzis1 , Maria Virvou1 and Lakhmi C. Jain2, 3
(1) Department of Informatics, University of Piraeus, 18534 Piraeus,
Greece
(2) Liverpool Hope University, Liverpool, UK
(3) University of Technology Sydney, Sydney, Australia
George A. Tsihrintzis (Corresponding author)

Email: geoatsi@unipi.gr
Abstract
The field of Machine Learning and its sub-field of Deep Learning are
most active areas of research in Artificial Intelligence, as researchers
worldwide continuously develop and announce both new theoretical
results and innovative applications in increasingly many and diverse
other disciplines. The book at hand aims at exposing its readers to
some of the most significant recent advances in Machine
Learning/Deep Learning-based technologies. At the same time, the
book aims at honouring Professor Nikolaos G. Bourbakis, an
outstanding researcher in this area who has contributed significantly to
the development of Machine Learning/Deep Learning-based
technologies. As such, the book is directed towards professors,
researchers, scientists, engineers and students in computer science-
related disciplines. It is also directed towards readers who come from
other disciplines and are interested in becoming versed in some of the
most recent progress in Machine Learning/Deep Learning-based
technologies. An extensive list of bibliographic references at the end of
each chapter guides the readers to probe deeper into their areas of
interest.
1.1 Editorial Note

The 4th Industrial Revolution is rising [1, 2], moving human civilization
into a new era and restructuring human societal organization into, so-
called, “Society 5.0” [3, 4]. One of the main driving forces is Artificial
Intelligence [5], in general, and Machine Learning [6], in particular. The
term “Machine Learning” dates back to 1959, as it was introduced in a
nominal paper by Samuel [7]. Today, Machine Learning has grown into
a multi-disciplinary approach of very active and intense research
worldwide [8–14], which aims at incorporating learning abilities into
machines. More specifically, the aim of Machine Learning research and
applications is to enhance machines with mechanisms, methodologies,
procedures and algorithms that allow them to become better and more
efficient at performing specific tasks, either on their own or with the
help of a supervisor/instructor.
Within the umbrella of Machine Learning, the sub-field of Deep
Learning (for example, see [15] for an-easy-to-follow first read on Deep
Learning) includes all Machine Learning methods based on artificial
neural networks, as inspired from biological neural structures. Today,
Deep Learning includes multi-layered neural processing paradigms,
such as convolutional neural networks, recurrent neural networks and
deep belief networks [15–18].
Due to its worldwide pace of formidable growth both in new
theoretical results and in achieving high scores in successful new
technological application areas, the field of Machine Learning/Deep
Learning has already had a significant impact on many aspects of
everyday life, the workplace and human relationships and is expected
to have an even wider and deeper impact in the foreseeable future [8–
19].
Specifically, some of the technological application areas, where the
use of Machine Learning/Deep Learning-based approaches has met
with success, include the following:
I. Machine Learning/Deep Learning in Socializing and Entertainment
II. Machine Learning/Deep Learning in Education
III. Machine Learning/Deep Learning in Security
IV. Machine Learning/Deep Learning in Time Series Forecasting
V. Machine Learning in Video Coding and Information Extraction
The book at hand aims at updating the relevant research

communities, including professors, researchers, scientists, engineers
and students in computer science-related disciplines, as well as the
general reader from other disciplines, on the most recent advances in
Machine Learning/Deep Learning-based technological applications. At
the same time, the book also aims at honouring Professor Nikolaos G.
Bourbakis, an outstanding researcher and educator, who has conducted
leading research in this field and has inspired, advised and mentored
tens of students, fellow researchers and colleagues.
More specifically, the book consists of an editorial chapter (Chap. 1)
and an additional ten (10) chapters. All chapters in the book were
invited from authors who work in the corresponding chapter theme
and are recognized for their significant research contributions. In more
detail, the chapters in the book are organized into five parts, as follows:
The first part of the book consists of two chapters devoted to
Machine Learning/Deep Learning in Socializing and Entertainment.
Specifically, Chap. 2, by F. Di Martino and S. Senatore, is entitled
“Semi-Supervised Feature Selection Method for Fuzzy Clustering of
Emotional States from Social Streams Messages.” The authors propose a
new method based on a fuzzy clustering algorithm that takes into
account human suggestions for feature selection to capture the user
mood in decision-making processes.
Chapter 3, by K. Karpouzis and G. Tsatiris, is entitled “AI in (and for)
Games.” The authors discuss some of the most common and widely
accepted uses of Artificial Intelligence/Machine Learning algorithms in
games and how intelligent systems can benefit from those, elaborating
on estimating player experience based on expressivity and
performance, and on generating proper and interesting content for a
language learning game.
The second part of the book consists of two chapters devoted to
Machine Learning/Deep Learning in Education.
Specifically, Chap. 4, by V. Zafeiropoulos and D. Kalles, is entitled
“Computer-Human Mutual Training in a Virtual Laboratory
Environment.” The authors discuss recent developments of Onlabs, an
interactive 3D virtual lab developed at the Hellenic Open University,
and assess its performance via two separate machine learning
techniques, namely a genetic algorithm and back-propagation on an
artificial neural network.
Chapter 5, by G. Kostopoulos and S. Kotsiantis, is entitled “Exploiting
Semi-supervised Learning in the Education Field: A Critical Survey.” The
authors provide a comprehensive review of the applications of Semi-
Supervised Learning in the fields of Educational Data Mining and
Learning Analytics. Their review indicates that Semi-Supervised
Learning constitutes a very effective tool for both early and accurate
prognosis of student learning outcomes when compared to traditional
supervised methods.
The third part of the book consists of two chapters devoted to
Machine Learning/Deep Learning in Security.
Specifically, Chap. 6, by M. Alamaniotis and A. Heifetz, is entitled
“Survey of Machine Learning Approaches in Radiation Data Analytics
pertained to Nuclear Security.” The authors provide a comprehensive
survey and discussion of Machine Learning and Data Analytics methods
pertaining to nuclear security and also discuss further trends and how
the data analytics can further enhance nuclear security by effectively
analyzing radiation data.
Chapter 7, by D. Gumusbas and T. Yildirim, is entitled “AI for
Cybersecurity: ML-Based Techniques for Intrusion Detection Systems.”
The authors discuss problems in cybersecurity and their potential
Machine Learning-based solutions and point to open avenues of future
research in this area.
The fourth part of the book consists of two chapters devoted to
Machine Learning/Deep Learning in Time Series Forecasting.
Specifically, Chap. 8, by A. Karanikola, C. M. Liapis and S. Kotsiantis,
is entitled “A Comparison of Contemporary Methods on Univariate Time
Series Forecasting.” The authors compare the performance of several
contemporary forecasting models that are considered state of the art,
including Autoregressive Integrated Moving Average (ARIMA), Neural
Basis Expansion Analysis (NBEATS), Probabilistic Time Series
Modeling, Deep Learning-based models and other.
Chapter 9, by S. A. A. Ojeda, E. C. Peramo, and G. A. Solano, is entitled
“Application of Deep Learning in Recurrence Plots for Multivariate
Nonlinear Time Series Forecasting.” The authors present a framework
for multivariate nonlinear time series forecasting that utilizes phase
space representations and deep learning.
Finally, the fifth part of the book consists of two chapters devoted
to Machine Learning in Video Coding and Information Extraction.
Specifically, Chap. 10, by A. Angeleas and N. G. Bourbakis, is entitled
“Formal and Statistical AI Tool for Complex Human Activity Recognition.”
The authors present a novel end-to-end Machine Learning-based tool
for complex human activity recognition and behavioral interpretation,
backed by formal and statistical information.
Finally, Chap. 11, by J. Li, M. Yang, Y. X. and Z. Li, is entitled “A CU
Depth Prediction Model Based on Pre-trained Convolutional Neural
Network for HEVC Intra Encoding Complexity Reduction.” The authors’
work is in the area of High Efficiency Video Coding and proposes a
hierarchical Coding Unit depth prediction model based on a pre-trained
convolutional neural network to predict the Coding Tree Unit split
pattern based on the image block.
1.2 Book Summary and Future Volumes
In this book, we have presented some significant advances in Machine
Learning/Deep Learning-based technologies, while honouring
Professor Nikolaos G. Bourbakis for his research contributions to this
discipline. The book is directed towards professors, researchers,
scientists, engineers and students in computer science-related
disciplines. It is also directed towards readers who come from other
disciplines and are interested in becoming versed in some of the most
recent advances in these active technological areas. We hope that all of
them will find the book useful and inspiring in their works and
researches.
The book has also come as a seventh volume, following six previous
volumes by the Editors devoted to aspects of various Machine Learning
Paradigms [8–10, 12, 14, 19]. As societal demand continues to pose
challenging problems, which require ever more efficient tools,
methodologies, systems and computer science-based technologies to be
devised to address them, the readers may expect that additional related
volumes will appear in the future.
References
1. J. Toonders, Data is the new oil of the digital economy. Wired. https://www.
wired.c om/insights/2014/07/data-new-oil-digital-economy/
2.
K. Schwabd, The Fourth Industrial Revolution—what it means and how to
respond. Foreign Affairs. https://www.foreignaffairs.c om/articles/2015-12-12/
fourth-industrial-revolution. Accessed 12 Dec 2015
3.
From Industry 4.0 to Society 5.0: the big societal transformation plan of Japan,
https://www.i-scoop.eu/industry-4-0/society-5-0/.
4.
Society 5.0, https://www8.c ao.go.j p/c stp/english/society5_0/index.html
5.
E. Rich, K. Knight, S.B. Nair, Artificial Intelligence, 3rd edn. (Tata McGraw-Hill
Publishing Company, 2010)
6.
J. Watt, R. Borhani, A.K. Katsaggelos, Machine Learning Refined—Foundations
Algorithms and Applications, 2nd edn. (Cambridge University Press, 2020)
7.
A. Samuel, Some studies in machine learning using the game of checkers. IBM J.
3(3), 210–229 (1959)
[MathSciNet][Crossref]
8.
A.S. Lampropoulos, G.A. Tsihrintzis, Machine learning paradigms—applications in
recommender systems, in Intelligent Systems Reference Library Book Series, vol.
92 (Springer, 2015)
9.
D.N. Sotiropoulos, G.A. Tsihrintzis, Machine learning paradigms—artificial
immune systems and their application in software personalization, in Intelligent
Systems Reference Library Book Series, vol. 118 (Springer, 2017)
10.
G.A. Tsihrintzis, D.N. Sotiropoulos, L.C. Jain (eds.), Machine learning paradigms—
advances in data analytics, in Intelligent Systems Reference Library Book Series,
vol. 149 (Springer, 2018)
11.
A.E. Hassanien (ed.), Machine learning paradigms: theory and application, in
Studies in Computational Intelligence Book Series, vol. 801 (Springer, 2019)
12.
G.A. Tsihrintzis, M. Virvou, E. Sakkopoulos, L.C. Jain (eds.), Machine learning
paradigms—applications of learning and analytics in intelligent systems, in
Learning and Analytics in Intelligent Systems Book Series, vol. 1 (Springer, 2019)
13.
J.K. Mandal, S. Mukhopadhyay, P. Dutta, K. Dasgupta (eds.), Algorithms in machine
learning paradigms, in Studies in Computational Intelligence Book Series, vol. 870
(Springer, 2020)
14.
M. Virvou, E. Alepis, G.A. Tsihrintzis, L.C. Jain (eds.), Machine learning paradigms
—advances in learning analytics, in Intelligent Systems Reference Library Book
Series, vol. 158 (Springer, 2020)
15.
J. Patterson, A. Gibson Deep Learning—A Practitioner’s Approach (O’ Reilly, 2017)
16.
Y. Bengio, A. Courville, P. Vincent, Representation learning: a review and new
perspectives. IEEE Trans. Pattern Anal. Mach. Intell. 35(8), 1798–1828 (2013)
[Crossref]
17.
J. Schmidhuber, Deep learning in neural networks: an overview. Neural Netw. 61,
85–117 (2015)
[Crossref]
18.
Y. Bengio, Y. LeCun, G. Hinton, Deep learning. Nature 521(7553), 436–444 (2015)
[Crossref]
19.
G.A. Tsihrintzis, L.C. Jain (eds.), Machine learning paradigms—advances in deep
learning-based technological applications, in Learning and Analytics in Intelligent
Systems Book Series, vol. 18 (Springer, 2020)
Part I
Machine Learning/Deep Learning in
Socializing and Entertainment
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022
G. A. Tsihrintzis et al. (eds.), Advances in Machine Learning/Deep Learning-based Technologies, Learning and
Analytics in Intelligent Systems 23
https://doi.org/10.1007/978-3-030-76794-5_2
2. Semi-supervised Feature Selection Method for

Fuzzy Clustering of Emotional States from Social
Streams Messages
Ferdinando Di Martino1 and Sabrina Senatore2
(1) Dipartimento di Architettura, Università degli Studi di Napoli Federico II, Via Toledo
402, 80134 Napoli, Italy
(2) Dipartimento di Ingegneria dell’Informazione ed Elettrica e Matematica applicata,
Università degli Studi di Salerno, Via Giovanni Paolo II, 132, 84084 Fisciano, Salerno,
Italy
Ferdinando Di Martino (Corresponding author)

Email: fdimarti@unina.it
Sabrina Senatore
Email: ssenatore@unisa.it
Abstract
Capturing the text content, especially when it reflects the human emotional states and
feelings, is crucial in every decision-making process: from the item purchase to the
marketing campaign, the user mood is becoming an essential peculiarity to always
monitoring. This work proposed a new method based on a fuzzy clustering algorithm that
takes into account human suggestions for feature selection. The method exploits two fuzzy
indices, namely, the feature relevance that is initially provided by the human expertise and
the feature incidence on a specific cluster. The Extended Fuzzy C-Means (EFCM) clustering
is used to balance the two “dueling” indexes; a t-norm operator-based feature importance
index enables the appropriate feature set selection. Experimental results on social
message streams show the method’s effectiveness in supporting those emotions the
human considers relevant in the textual context.
2.1 Introduction
Nowadays, high throughput technologies routinely produce large data that are recorded
and stored for analytics purposes. In the Social Web particularly, the continuous user-
generated content needs to be arranged appropriately to accelerate text analysis and
information retrieval tasks. Data often contain irrelevant and redundant features, with a
high level of noise.
Especially in the classification task, large feature vectors could significantly slow down
the process, and, even though such vectors are expected to have more discriminating
power, practically, they often produce models that do not reflect a quite generalized data
representation.
This problem is quite evident in the processing of textual information, such as papers,
websites, reviews, twitters, or snippets. The expressiveness of natural language
emphasizes the difficulty to discriminate appropriate features that support accurately the
classification methods. On the other hand, the increasing volume of opinionated data
disseminated on the Web needs enhanced approaches to analyze and process data in an
efficient way, capturing the actual meaning behind the text. The natural language indeed is
imprecise and ambiguous and, in general, the text is composed of a loosely structured
sequence of words and symbols that can support humans in capturing the actual meaning
of the sentences but, this activity is quite complex for the computational systems that
could not infer a right context for a group of words. These issues can be amplified if text
mining activities are targeted at capturing the emotions and the sentiments from opinions
[20].
Analyzing user-generated content in social media to capture people’s emotions and
understand public attitude and mood is a crucial task for market analysis, business,
political consensus study. Consumers can influence other users’ consumption activities: an
opinion, a comment, a reaction quickly reach global audiences who share similar interests
in a product or brand. Text Mining is a complex activity that needs to discover relevant
information from a large collection of textual information, which is often unstructured,
redundant, and duplicated.
Feature selection becomes a mandatory preprocessing phase to reduce the
dimensionality and eliminate the duplication and unwanted features in the data.
Many feature selection algorithms have been developed in literature [7], often studied
as optimization problems [9]. In Information Retrieval (IR) approaches, the Bag-of-Words
(BoW) is the most known vector space model used to represent documents. It is filled by
word frequencies over a fixed dictionary. The feature selection methods remove the lowest
ranking terms based on a scoring function, such as term occurrences [16] and frequencies,
TF-IDF [17], as well as mutual information (MI) and chi-squared ranking (χ2) [12].
Selecting the highest-ranked terms does not guarantee to get the most relevant feature,
especially in the text mining tasks, where polysemy and synonymy can affect the
classification tasks: redundant features do not contribute to adding new information to
describing the concept, as well as irrelevant features can simply add noise to the mining
process [20]. On the other hand, the reduction of the feature set impacts the size of the
data space, and therefore also decreases the complexity of the classification and prediction
problems.
In addition to the traditional Information Retrieval and Text Mining approaches, which
are mainly based on a preliminary feature set definition, many approaches in the literature
aim at seeking latent semantics from the data space to face polysemy, synonyms,
homonyms, phrases dependencies issues. Latent models [1] are useful to identify the
semantic concepts in text documents and uncover the latent semantic structure embedded
in document collections. In [5] latent models help discriminating emotions from textual
space: the projection of words expressing sentiments and emotions in the same space
topology provides effective contexts to overcome linguistic ambiguity in the natural
language. Sentiment Analysis has been extensively studied in recent years [11–19], often
focusing on natural language processing techniques to face all the issues related to
understanding the written language targeted at interpreting human moods.
In [1], the Latent Dirichlet Allocation (LDA) model is used to extract latent topics and is
combined with a Bayesian approach to extract concepts to associate with the topics. Some
approaches [8–21] adopt external resources such as WordNet [22] as well as dictionaries,
thesauri, and knowledge bases to discriminate sense and the context of terms in sentences.
Classification in natural language processing tasks finds in Deep Learning techniques
[14] compelling methods to capture the complexity of language, overcoming problems,
such as the curse of dimensionality, since the linguistic text was represented with sparse
matrices (high-dimensional features). With the recent popularity of word embeddings,
neural-based approaches exhibit good performance compared to more traditional machine
learning models. Empirical evidence shows that discovering linguistic patterns remain an
open issue in language understanding, due to the complexity of natural language, that
through metaphors, rhetoric, figurative expressions make ineffective known automatic
models for feature extraction and selection.
This paper presents a novel feature selection method applied to fuzzy clustering. The
algorithm is called FS-EFCM (Future Selection on Extended Fuzzy C-Means) and extends
the EFCM algorithm [10]. The algorithm takes external scores into account as additional
parameters for the initial configuration. The idea is to allow human suggestion in the
discrimination of relevant features, viz., the features that are crucial to describe the
domain of interest the features are from, and then to affect the clustering process into
discarding irrelevant and noise information.
Thus, experts can provide their relevance values (weights) to each feature based on
their expertise and knowledge of the reference domain.
During the EFCM execution, some features could be discarded based on the expert
feature relevance selection. The features are also evaluated with respect to their impact,
i.e., the affecting on the formation of the clusters. Both feature relevance and incidence are
monitored during the FS-EFCM execution, to evaluate which features are crucial for both
the clustering performance and the experts.
The remainder of the paper is organized as follows. Section 2.2 introduces the
proposed FS-EFCM algorithm: a general overview of the algorithm is presented firstly, then
the main steps and the pseudocode describe the algorithm in detail. Additional
investigation on the parameter setting configuration is also provided. Section 2.3 is
devoted to the experimental results. A dataset composed of tweet streams is analyzed to
classify the tweet trends by capturing the sentiments and emotions from text analysis. The
experiments show the effectiveness of the proposed method. Finally, conclusions will be
given in the last section.
2.2 The FS-EFCM Algorithm

The FS-EFCM algorithm accomplishes data fuzzy clustering by introducing a feature
selection method that filters irrelevant features out. It is from an extension of the EFCM
[10], a partitive fuzzy clustering algorithm that, based on the well-known FCM algorithm
[4, 3], overcomes its drawbacks, such as a priori choice of the number of clusters and the
sensibility to the presence of noise and outliers. Additionally, EFCM is robust to the
partition initialization and does not require to validate the produced clustering
partitioning over several random initializations.
Figure 2.1 sketches the main steps of the FS-EFCM algorithm, showing the EFCM
module embedded. During the EFCM runs, some features are candidates to be discarded,
since they are not meaningful in the feature set and, at the same time, they do not affect
the cluster formation. The EFCM is then re-executed until the stability condition is not
satisfied.
Fig. 2.1 Logical overview of the algorithm
The algorithm takes as input the data collection and the expert-driven scores (weight)
associated with the features. Depending on the data, plausible data analysis and
preprocessing activities could be started to make it suitable for processing by the
algorithm. For example, the textual dataset must be processed by apply typical NLP tasks
(i.e., tokenization, stemming, stop-word removal, pos tagging, etc.), sentiment, and
emotion analysis instead focuses on capturing the emotional aspect embedded in the word
or sentence meaning. The EFCM algorithm works on data translated in a matrix form.
Each score given by the human experts and associated with a feature describes how
that feature is relevant in the domain of interest, according to the expert viewpoint.
As shown in Fig. 2.1, the scores are processed (Feature relevance (FR) Estimation) to
rescale them according to the appropriate range and evaluation metrics.
Acquired the input, the algorithm implements an iterative process: in each iteration,
the EFCM is launched and, until the stability condition is not verified, the whole algorithm
is re-run by updating the parameter configuration. Precisely, the EFCM output is targeted
at evaluating the incidence of each feature in the clustering formation (Feature Incidence
(FI) Assessment).
The stability condition is strictly correlated to two important indices of the algorithm:
the feature relevance FR and the feature incidence FI, which represents the importance of a
feature from the human viewpoint and the incidence of the same feature from the data
distribution in the clustering structure, respectively.
The algorithm stops when a condition of stability is reached, i.e., when all the features
are strongly affecting the clustering formation.
Otherwise, when the stability condition is not satisfied, a further evaluation based on
the two indices individuates the features candidate to be discarded. Once removed, the
process is re-iterated con the remaining features.
2.2.1 EFCM Execution: Main Steps
The FS-EFCM algorithm can be described by the following steps:
1. Feature relevance (FR) Estimation: the collected score assigned by experts to the
features sh with h = 1, …, H are translated in a proper scale in the range [0, 1]. In
general, scores assigned by experts could be defined in a scale correlated to the data
domain. Thus, an index could be necessary to bound the feature score in the interval
[0, 1]. For example, the score sh could be fuzzified assigning a membership degree
μFR(sh) to a pre-defined fuzzy set (e.g., a sigma fuzzy set on a universe of a discourse
given by an interval of the real line).
2. EFCM algorithm execution once given the data and feature scores, the EFCM algorithm
is executed; in the first run, all the input features are used. The generated clusters are
hyperspheres in the feature space.
3. Feature incidence (FI) assessment: the clusters generated by the EFCM are analyzed:
the incidence of each feature in the clustering structure is calculated, by evaluating the
hth feature component impact on each cluster prototype (measured as the distance of
the feature components between the cluster prototype pairs). More formally, at the tth
algorithm iteration, the weight value w(t)h of hth feature component, is feature
incidence value and it is calculated as follows:
(2.1)
where v′ih and v′kh are the hth component values of all the cluster center pairs,
evaluated at the tth iteration. The weight value w(t)h assumes values between 0 and 1;
the higher the value is, the more the feature affects the cluster formation. Similarly to
the FR index, the incidence value is used to calculate the corresponding membership
degree to a prefixed fuzzy set.
4. Stability Condition Check: The algorithm stops when the difference between the FI
values of a feature in two successive iterations is below a prefixed threshold θ.
Formally, the stability condition is given by:
(2.2)
where μFI(wh) is in general the membership degree to a fuzzy set defined on the
feature wh. If the condition holds, the current feature keeps being part of the feature
set, otherwise the algorithm continues to the next step.
5. Discard less significant features: the remaining features are the candidate to be
removed by the feature set since their contribution might be not significant in the
clustering process. The selection of features candidate to be removed from the feature
set is achieved by defining a new measure of significance μh of the hth feature for the
clustering structure, applying the t-norm operator as follows:
(2.3)
If μh is lower than a prefixed threshold δ, then the hth feature is unmeaningful, and
then it is removed from the current feature set. Finally, the process returns to step 2,
considering just the filtered features in the next iteration.
The pseudocode of the FS-EFCM is shown Listing 1.
Listing 1. Algorithm: FS-EFCM
2.2.2 Initial Parameter Setting

The parameter δ represents the threshold value below which a feature could be removed
from the feature set. Its setting is crucial to controls the enhancement of the algorithm
execution.
Thus, a preliminary step has been defined to set the proper value of the parameter δ.
The rationale behind this step is to identify potential removable features, even though they
were considered relevant by the experts.
Let nR be the number of features considered relevant by the experts and nD is the
number of the candidate features to be discarded after the EFCM execution. If nD is greater
than 50% of nR (i.e., many features that are relevant for the expert likely will be discarded),
then the initial value of the parameter δ is re-set to a lower value.
Listing 2 shows the pseudocode of the algorithm SelectDeltaThreshold describing how
the parameter δ is refined. Initially, nR is calculated: it is the number of features considered
very relevant by the experts, i.e., those with membership degree μFR ≥ 0.7 (lines 6–8). The
value 0.7 is selected arbitrarily and it guarantee to select just the more meaningful
features.
The relevant features candidate to be removed nD are stored in the array sig[] of nD
size: lines 12–19 indeed describe how the array is filled. Line 13 shows the condition for
discarding a feature: like in FS-EFCM Algorithm, μh of Eq. (2.3) must be lower than δ and,
at the same time, checks if the feature is relevant (μFR ≥ 0.7). The feature significance value
is added to the array sig[], if both conditions hold. Then, the array is ranked ascendingly, by
feature significance values. The (non-zero) value in the intermediate (or the next non-zero
intermediate) position of the array will be the new value of the threshold δ.
Let us notice that the value of the parameter δ must be lower than the initial value
μFR(sh), the membership degree assigned by the expert for the hth feature, and calculated
for the FR fuzzy set. This guarantee that line 15 holds and further refinement of δ is
applied in line 20.
This strategy ensures that more than half of the features relevant for the experts are
not removed from FS-EFCM initially.
Listing 2. Algorithm: SelectDeltaThreshold
A further choice concerns the t-norm operator used in Line 13 of Listing 2 and defined
in Step 5 of the FS-EFCM algorithm (Sect. 2.2.1). Among the families of t-norms defined in
literature, the most widely used in application problems are the triangular norms, shown
as follows:
– minimum (Gödel) t-norm x● y = min(x, y)
– product (Goguen) t-norm x● y = x⋅y
– Lukasiewicz t-norm x● y = max(x + y − 1, 0)
Depending on the selected t-norm, different fuzzy intersections are generated; in
particular, the minimum t-norm is the most used in fuzzy controls, whereas the product t-
norms produces a more drastic intersection than the minimum t-norm.
2.3 Experimental Results

2.3.1 Dataset
The experiments have been carried out on a Twitter dataset.
Twitter is one of the most preferred social networks, where people express their
emotions, opinions about events, interests, items. It offers a comfortable platform for
chatting, reading tweets, reacting accordingly by writing their comments through tweets,
or sharing tweets written by others.
The dataset is composed of four hundred thousand public tweets posted by users from
May 2018 to July 2018 in the cities of Washington, New York, and London [13]. All the
tweets are in English and can have one or more hashtag. Tweets with the same hashtag are
grouped in a unique document file: the hashtag allows the topic to be well-characterized.
Files get as a name the hashtag that mainly appears in the collected tweets. Tweets
without hashtag are discarded because they cannot be associated with a document.
The tweet collection has been preprocessed, applying elementary NLP tasks. Precisely,
data have been stemmed, i.e., each word has been reduced to its inflectional form; then the
stop words have been removed as well. Syntactic slangs and non-conventional dictionary
terms are also discarded.
To apply the FS-EFCM algorithm, the data should be a term-documents matrix: each
hashtag-based document is represented as a vector, and each cell contains a numeric value
associated with a feature.
2.3.2 Feature Selection

In the text mining approaches, the feature set is usually composed of a collection of terms
describing the data space peculiarities; often they are extracted by applying some metrics
on the text (e.g., tf-idf-ranked top-k terms).
This experimentation aims at capturing emotions and human behavior from the tweet
trends. From tweets analysis, it is possible to inspect people’s sensations and feelings on
local or world events; to investigate the main emotions and opinions that justify some
behaviors.
Sentiments and opinions indeed are concealed in the sentences, typically associated
with adjectives and verbs; then the intrinsic meaning of some textual expressions is not
amenable to rigid linguistic patterns [2].
To this purpose, the feature set should be composed of a representative word set that
describes the most relevant emotions.
A dataset1 composed of sixty emotional categories (target classes) has been selected to
cover a wide range of relevant emotional categories. A half of these categories describe
pleasant feelings (they are labeled “open”, “happy”, “alive”, “good”, “love”, “interest”,
“positive”, “strong”) whereas the remaining eight categories are related to
difficult/unpleasant feelings (“angry”, “depressed”, “confused”, “helpless”, “indifferent”,
“afraid”, “hurt”, “sad”). Each emotional category is in turn composed of a set of words that
“explain” the category; these words are generally synonyms or terms with similar meaning
(e.g., “happy” and “glad”). The features set is comprehensively composed of 250 features,
labeled with the words in the emotional categories.
2.3.3 FS-EFCM at Work
The FS-EFCM takes as input a matrix composed of 250 columns (term-features) and 5607
rows (documents extracted from a data stream of about four hundred and fifty thousand
tweets, as stated above).
Each entry of the data point associated with a document vector is calculated by using
the TF-IDF metrics.
For completeness, in Table 2.1, the name of all the features of the data points in our
dataset.
Table 2.1 Features of the documents in our emotional dataset
Abomin Absorb Accept Ach Admir Affect Affection Afflict Afraid

Aggress Agon Alarm Alien Aliv Alon Amaz And Angri
Anguish Anim Annoy Anxious Appal Asham Attract Bad Bitter
Bless Boil Bold Bore Brave Bright Calm Certain Challenge
Cheer Clever Close Cold Comfort Concern Confid Confus Consider
Content Courag Coward Cross Crush Curious Dare Deject Delight
Depress Depriv Desol Despair Desper Despic Determin Detest Devot
Diminish Disappoint Discourag Disgust Disillus Disinterest Dismay Dissatisfi Distress
Distrust Domin Doubt Drawn Dull Dynam Eager Earnest Eas
Easi Ecstat Elat Embarrass Empti Encourag Energet Engross Enrag
Enthusiast Excit Fascin Fatigu Fear Festiv Forc Fortun Free
Frighten Friski Frustrat Fume Gay Glad Gleeful Good Great
Grief Griev Guilti Happi Hardi Hate Heartbroken Helpless Hesit
Hope Hostil Humili Hurt Import Impuls Incap Incens Indecis
Indiffer Indign Inferior Inflam Infuri Injur Inquisit Insensit Inspir
Insult Intent Interest Intrigu Irrit Joyous Jubil Keen Kind
Liber Lifeless Lone Lost Lousi Love Lucki Menac Merri
Miser Misgiv Mourn Nervous Neutral Nonchal Nosi Offend Offens
Open Optimist Overjoy Pain Panic Paralyz Passion Pathet Peac
Perplex Pessimist Play Pleas Posit Powerless Preoccupi Provoc Provok
Quak Quiet Re-enforc Reassur Rebelli Recept Reject Relax Reliabl
Repugn Resent Reserv Restless Sad Satisfi Scare Secur Sensit
Seren Shaki Shi Skeptic Snoopi Sore Sorrow Spirit Stew
Strong Stupefi Sulki Sunni Sure Surpris Suspici Sympathet Sympathi
Tear Tenaci Tender Tens Terribl Terrifi Thank Threaten Thrill
Timid Torment Tortur Touch Toward Tragic Unbeliev Uncertain Understand
Uneasi Unhappi Uniqu Unpleas Unsur Upset Useless Victim Vulner
Wari Warm Weari Woeful Wonder Worri Wrong
To each data point (or document), one of the sixteen emotional categories is associated
as the target class. This classification is obtained applying the method described in [6].
Initially, the experts assigned to each feature a relevance score in the range [1, 10]. In
the initial configuration for running FS-EFCM, the threshold δ is set to 0.1, and the stop
iteration/convergence threshold θ is set to 0.01.
Furthermore, we set the EFCM fuzzifier parameter to 2, the initial number of clusters to
50, the EFCM merging threshold to 0.01 and the EFCM stop iteration threshold to 0.001.
In the preprocessing phase, the SelectDeltaThreshold algorithm is executed to select the
best value of δ.
After its execution, the significance of 135 features is below the prefixed threshold δ;
just 20 of them are considered relevant for the expert (with a relevance index greater than
or equal to 0.7). Then, the new value of δ is set to 0.05, and the SelectDeltaThreshold
algorithm is re-run. 96 features, 9 of which are relevant to the expert, are removed. Table
2.2 summarizes these results.
Table 2.2 Results obtained in the preprocessing phase
Number of Threshold Features Relevant features New Feature Features

features δ potentially potentially threshold δ removed selected
removable removable
250 0.1 135 20 0.05 96 154
Then, EFCM runs on the remaining 154 selected features. The algorithm stops after 8
cycles/iterations. In the last three cycles, the number of features decreases to 20, keeping
this value until the last cycle. Table 2.3 shows the results obtained after each algorithm
iteration.
Table 2.3 Results obtained launching FS-EFCM
Cycle Number of Number of Feature Features Stop iteration

features clusters removed selected value
1 154 20 62 92 0.42
2 92 18 37 55 0.23
3 55 16 22 33 0.15
4 33 16 9 24 0.09
5 24 16 3 21 0.05
6 21 16 1 20 0.02
7 20 16 0 20 0.013
8 20 16 0 20 0.005
After the eighth cycle the stop iteration value is less than the threshold θ and the
algorithm stops.
In Table 2.4, the remaining twenty features and the measure of their significance are
shown.
Table 2.4 Significances of the final selected features
Feature name Feature significance
Admir 0.09
Afraid 0.15
Anxious 0.18
Attract 0.11
Bad 0.16
Good 0.16
Great 0.13
Happi 0.15
Hurt 0.14
Love 0.10
Pain 0.21
Panic 0.12
Passion 0.15
Sad 0.13
Scare 0.08
Thank 0.22
Touch 0.14
Unhappi 0.15
Upset 0.13
Worri 0.08
The final number of clusters is 16. The final document membership is calculated,
assigning it to the cluster to which the document belongs with the highest membership
degree.
Hence, cluster-class mapping is achieved by associating each cluster with the class to
which most of the documents that have been assigned to the cluster belong to.
Table 2.5 shows for each emotional category, the number of documents in the
class/category, the cluster label associated with the class, the number of documents
assigned to this cluster, and the number of documents assigned to the cluster that belong
to the class (well-classified documents).
Table 2.5 Number of documents assigned to the class and assigned to the correspondent cluster
Class Documents belonging to Cluster Documents assigned to the Well classified

the class cluster documents
Happy 1380 C1 1369 1159
Open 366 C2 380 301
Confused 24 C3 31 13
Strong 55 C4 61 36
Interested 55 C5 49 35
Depressed 49 C6 41 26
Positive 421 C7 430 334
Hurt 377 C8 369 296
Good 428 C9 435 337
Alive 187 C10 182 141
Helpless 164 C11 166 128
Love 1186 C12 1175 1007
Angry 434 C13 426 352
Afraid 315 C14 322 257
Sad 161 C15 167 131
Indifferent 5 C16 4 3
About 81% of documents belonging to a class are assigned to the right cluster. The
performance of FS-EFCM measured calculating the accuracy, precision, recall, and F1-score
classification indices are shown in Table 2.6.
Table 2.6 Classification performance indices
Measure Value (%)

Accuracy 77.02
Precision 77.88
Recall 77.21
F1-score 77.54
Now, let us consider just those documents whose membership degree to the cluster to
which they are assigned is high or higher than a specific threshold; let us set that threshold
to 0.6.
The number of documents whose membership degree to the assigned cluster is greater
than 0.6 is 2047, which means it is about 37% of all the document collection.
Table 2.7 shows statistics similar to Table 2.5 but considering just the documents that
strongly belong to the cluster they are assigned to.
Table 2.7 Number of documents assigned to the class and assigned to the correspondent cluster
considering the documents with membership degree to the assigned cluster greater than 0.6

Happy 497 C1 494 487
Open 134 C2 133 128
Confused 10 C3 10 9
Strong 18 C4 20 17
Interested 24 C5 23 22
Depressed 21 C6 23 20
Positive 148 C7 149 143
Hurt 122 C8 119 116
Good 157 C9 158 153
Alive 75 C10 75 72
Helpless 56 C11 57 53
Love 450 C12 452 446
Angry 160 C13 162 157
Afraid 116 C14 114 110
Sad 57 C15 56 52
Indifferent 2 C16 2 2
Let us notice that the percent of documents belonging to a class and assigned to the
correspondent cluster increase considerably (about 98%). Table 2.8 shows the
classification performance for this document selection, evidencing notable improvements:
all the metrics values are higher than 98%.
Table 2.8 Classification performance indices calculated for the documents in Table 2.7
Measure Value (%)

Accuracy 98.01
Precision 98.38
Recall 98.08
F1-score 98.23
These results highlight the effectiveness of the FS-EFCM, especially on supporting the
classification of social data drawing relevant emotion-driven user behavior.
The experimentation reveals how the approach can generate accurate clusters, well-
described by the features associated with each cluster. In fact, the final features are those
that better represent the data distribution in the clusters and give a clear idea about the
main sentiments and emotions expressed in the tweet trends analyzed.
2.4 Conclusion
This paper presents a semi-supervised clustering to classifying user generated. This paper
presents a semi-supervised clustering to classifying user-generated content from the
analysis of the main emotions expressed in the text.
The algorithm acquires from human experts some scores to assign a relevance degree
to the features in the reference domain. Our experimentation focused on capturing
emotions from tweet streams. For these reasons, the experts scored words expressing
emotions or sentiments (such as “joy”, “beautiful”, etc.). The FS-EFCM algorithm reaches a
trade-off between the feature relevance score provided by the experts and the feature
impact on the cluster formation.
The performance of the algorithm revealed not just the effectiveness of the approach,
but also its reliability in filtering very relevant features that clearly characterize the final
clusters.
Future development of the algorithm aims at investigating deeply how the documents
are associated with emotional categories. The idea is to consider not just the cluster to
which the document belongs, with the highest membership degree, but also the other
clusters to which the document can belong (with lower membership degree), to discover
which sets of multiple emotional categories emerge from the document.
References
1. L. Alsumait, D. Barbara, C. Domeniconi, On-Line LDA: adaptive topic models for mining text streams
with applications to topic detection and tracking, in Conference: Proceedings of the 8th IEEE
International Conference on Data Mining (ICDM 2008), Pisa, Italy (2008), 10 pp. https://doi.org/10.1109/
ICDM.2008.140
2.
L. Barbosa, J. Feng, Robust sentiment detection on twitter from based and noisy data, in Proceedings of
the 23rd International Conference on Computational Linguistics: Posters. COLING’10, Beijing, China
(Association for Computational Linguistics, Stroudsburg, PA, USA, 2010), pp. 36–44
3.
J.C. Bezdek, Pattern Recognition with Fuzzy Objective Function Algorithms (Kluwer Academic Publishers,
Norwell, MA, USA, 1981). https://doi.org/10.1007/978-1-4757-0450-1
4.
J.C. Bezdek, R. Ehrlich, W. Full, The fuzzy C-means clustering algorithm. Comput. Geosci. 10(2–3), 191–
203 (1984)
[Crossref]
5.
E. Cambria, Affective computing and sentiment analysis. IEEE Intell. Syst. 31(2), 102–107 (2016)
6.
D. Cavaliere, S. Senatore, V. Loia, Context-aware profiling of concepts from a semantic topological space.
Knowl. Based Syst. 130, 102–115 (2017)
7.
G. Chandrashekar, F. Sahin, A survey on feature selection methods. Comput. Electr. Eng. 40(1), 16–28
(2014). https://doi.org/10.1016/j .c ompeleceng.2013.11.024.ISSN 0045-7906
8.
K. Dave, S. Lawrence, D.M. Pennock, Mining the peanut gallery: opinion extraction and semantic
classification of product reviews, in Proceedings of the 12th International Conference on World Wide Web,
WWW ’03 (ACM, New York, NY, USA, 2003), pp. 519–528. https://doi.org/10.1145/775152.775226
9.
I. Guyon, A. Elisseeff, A., An introduction to variable and feature selection. J. Mach. Learn. Res. 3(2003),
1157–1182 (2003)
[zbMATH]
10.
U. Kaymak, M. Setnes, Fuzzy clustering with volume prototype and adaptive cluster merging. IEEE
Trans. Fuzzy Syst. 10(6), 705–712 (2002)
11.
B. Liu, Sentiment analysis and opinion mining. Synth. Lect. Human Lang. Technol. 5(1), 168 (2012).
https://doi.org/10.2200/S00416ED1V01Y201204HLT016
12.
C.D. Manning, P. Raghavan, H. Schutze, Introduction to Information Retrieval (Cambridge University
Press, New York, 2008), 506 pp. ISBN 978-0521865715
13.
F. Martino Di, S. Senatore, S. Sessa, A lightweight clustering-based approach to discover different
emotional shades from social message streams. Int. J. Intell. Syst. 1, 19 (2019). https://doi.org/10.1002/
int.22105
14.
D.W. Otter, J.R. Medina, J.K. Kalita, A survey of the usages of deep learning for natural language
processing. IEEE Trans. Neural Netw. Learn. Syst. (2020). https://doi.org/10.1109/TNNLS.2020.
2979670
15.
S. Poria, A. Gelbukh, A. Hussain, N. Howard, D. Das, S. Bandyopadhyay, Enhanced senticnet with affective
labels for concept-based opinion mining. IEEE Intell. Syst. 28(2), 31–38 (2013). https://doi.org/10.
1109/MIS.2013
16.
A. Rehman, J. Kashif, H.A. Babri, S. Mehreen, Relative discrimination criterion—a novel feature ranking
method for text data. Expert Syst. Appl. (Elsevier) 42, 3670–3681 (2015)
[Crossref]
17.
G. Salton, C. Buckley, Term-weighting approaches in automatic text retrieval. Inf. Process. Manag. 24(5),
513–523 (1988). https://doi.org/10.1016/0306-4573(88)90021-0
[Crossref]
18.
L. Stanchev, Semantic document clustering using a similarity graph, in 2016 IEEE Tenth International
Conference on Semantic Computing (ICSC) (2016), pp. 1–8. https://doi.org/10.1109/I CSC.2016.8
19.
C. Strapparava, R. Mihalcea, Learning to identify emotions in text, in Proceedings of the 2008 ACM
Symposium on Applied Computing, SAC’08 (ACM, New York, NY, USA, 2008), pp. 1556–1560. https://doi.
org/10.1145/1363686.1364052
20.
S. Sun, C. Luo, J. Chen, A review of natural language processing techniques for opinion mining systems.
Inf. Fusion 36, 10–25 (2017). https://doi.org/10.1016/j .inffus.2016.10.004
[Crossref]
21.
T. Wei, Y. Lu, H. Chang, Q. Zhou, X. Bao, A semantic approach for text clustering using WordNet and
lexical chains. Expert Syst. Appl. 42(4), 2264–2275 (2015). https://doi.org/10.1016/j .eswa.2014.10.023
[Crossref]
22.
G. Miller, C. Fellbaum, WordNet. An electronic lexical database. Cambridge, MA: MIT Press; (1998). 423
pp, ISBN: 978-0262061971.
Footnotes
1 www.enchantedlearning.c om.
Another random document with
no related content on Scribd:
The Project Gutenberg eBook of A tragikum
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.
Title: A tragikum
Author: Zsolt Beöthy
Release date: July 29, 2022 [eBook #68633]
Language: Hungarian
Original publication: Hungary: Kisfaludy-Társaság, 1885
Credits: Albert László from page images generously made available

by the Google Books Library Project
*** START OF THE PROJECT GUTENBERG EBOOK A TRAGIKUM

***
Megjegyzés:
A tartalomjegyzék a 630. oldalon található.
A TRAGIKUM.
ÍRTA
BEÖTHY ZSOLT.
KIADJA A KISFALUDY-TÁRSASÁG.
BUDAPEST
FRANKLIN-TÁRSULAT
MAGYAR IROD. INTÉZET ÉS KÖNYVNYOMDA.
1885.
FRANKLIN-TÁRSULAT NYOMDÁJA
A TRAGIKUM ELMÉLETE ÉS
JELENSÉGEI.
I.
A tragikum elemei.
Az életnek és művészetnek minden tragikai jelensége, végső

benyomásában, megnyugtató. Csak végső és határozó
benyomásában, mert teljes hatását egyetlen érzésbe foglalni ép oly
lehetetlen, mint általában a fenség képeiét, melyek a megaláztatás
érzelmeiből emelnek föl bennünket a végetlen eszméjéhez. A
tragikai jelenségek is izgató küzdelmeken, félelmes
összeütközéseken és részvétkeltő csapásokon át törekszenek
megnyugtatásunkra és fölemelésünkre. Hatásuk ebben éri
természetes tetőpontját s ebben hangzik el.
És mivel emelnek föl oda, a hol lelkünk felzaklatott érzései
lecsöndesűlnek s helyöket bizodalom, kibékülés, megnyugvás, sőt
gyönyör foglalja el? Oly eszmék érvényre juttatásával, győzelemre
vezetésével, melyeket a természeti és erkölcsi világban jogosan és
szükségszerűen uralkodóknak tekintünk; erőnk fentartóinak, hitünk
biztosítékainak, létünk alapjainak érzünk és tudunk. Legalább a
világnak arra a képére nézve, melyet a tragikus költő vagy művész
előttünk feltár, ilyenekűl ismerünk és elfogadunk. Sophokles
Antigone bukásával a közrend, Kreonéval a kegyelet eszméjét vezeti
diadalra. Romeo és Juliában a családi kötelék erejének, Julius
Caesarban a korszellem megmásíthatatlanságának, Coriolanban a
haza iránti hűség megtörhetetlenségének gondolata foglalja el
lelkünket. V. László halálával az eskü szentségének eszméje ül
diadalt, Bánk bán romlásában a törvényes hatalomé. Az Ember
tragédiája a végtelenben való részességünk éreztetése által békít ki
sorsunkkal. Mindezek az eszmék: haza és család, állami rend és
korszellem, természeti kényszer és erkölcsi törvény, szabadság és
gondviselés: egy közös fogalomra utalnak bennünket, melynek
öszhangos tartalmát képezik. Nem magukban álló, külön-külön
hatalmak, hanem csak elemei, töredékei, nyilatkozásai egy közös és
egyetemes hatalomnak, egy minden irányú, általános
tökéletességnek, mely uralkodik fölöttünk, vezeti lépéseinket,
irányozza tetteinket s kifejezi magát pályánkban. Lelkünket kibékülés
és megnyugvás szállja meg, mint Ádámét, a hányszor csak érezteti
magát velünk. Kénytelenek vagyunk megnyugodni benne, s
örömmel teszszük, mert csak ez biztosíthatja boldogulásunkat.
Érvényesűl általunk, igyekezeteink, cselekvéseink, életünk által, akár
szép szerével, akár törés-szakadás útján. Hódolatunk által, ha egész
teljességében lelkünkbe fogadva követjük; bukásunk által, ha
összeütközésbe kerűlünk vele. Ott Hunyadi János, itt Coriolan; ott
Regulus, itt V. László. Győzelme az utóbbi esetekben az egyes
emberrel való bizonyos ellentétességből, összetűzésből s ennek
bukásából ered: a diadal zászlaja romok fölött leng. Az olynemű
ütközés, mely e mozzanatokat kifejezésre juttatja, keletkezése,
folyama és kimenetele az, a mit tragikumnak szoktunk nevezni.
De hát az az eszme, melyet az egyetemessel szembe kerűlt
ember képvisel, teljes és kizáró ellentéte-e annak, a mely legyőzi őt?
Az általános tökéletességgel a teljes tökéletlenség tűz-e össze?
Néhány közönségesen ismert példa meggyőz az ellenkezőről.
Othello, a veronai szeretők, Ágnes asszony épen nem a
szeretetlenségnek, a szív megtagadásának képviselői s Ádám
mindentől távolabb van, semhogy az Úr mindenhatóságával
szemben a tehetetlenséget jelentse. Tarthatja-e valaki Toldit, kinek a
férfi-becsülettel össze nem férő álarczos játéka juttatja magát és
Rozgonyi Piroskát romlásra, férfiatlannak és lovagiatlannak? Sőt
ellenkezőleg: a férfias egyenesség példájáúl idézhetjük őt ép úgy,
mint a hogy Romeóban és Juliában a szeretőnek, Antigoneban a hű
testvérnek, mind a két Brutusban a föláldozó, tiszta keblű hazafinak
eszményét keressük. A mi mindennapi életünk kiváló tüneményeit,
kik ez eszmények felé közeledni látszanak, hasonlatúl az ő nevökkel
illetjük. Nincs ez áldozatra szánt küzdők között egy sem, kiben
bizonyos kiválóságot, emberi tökéletességet ne találnánk: itt az
érzésnek, ott az értelemnek, amott az akaratnak teljességét és
rendkívüliségét. E kiválóság nélkül nem lehet tragikai küzdelem.
Épen ez a teljesség, rendkívüliség, tökéletesség az, a mely egyfelől
harczra csábítja őket az egyetemes hatalom ellen, másfelől képesíti
rá.
Azonban ha ez úgy van, mikép érthető akkor a harcz közöttük?
Ha az is megeshetik, hogy az eszme, melyet ők képviselnek, s az, a
mely diadalmaskodik fölöttük, egy és ugyanaz: hogyan keletkezhetik
az összeütközés? Úgy, hogy kiválóságuk mindig valamely
gyarlósággal, tökéletességök bizonyos tökéletlenséggel párosúl
vagy nyilatkozik. E tökéletlenség, mely a veszedelmet rájok zúdítja,
állhat lelki életök egyoldalú fejlettségében,
egyensúlyozatlanságában, indulataik túlságos elhatalmasodásában.
Mindezt még végzetesebbé teheti rájok nézve helyzetök
elszigeteltsége vagy visszássága: az emberi tökéletlenségnek
szintén egy oldala. Antigone: igazi eszménye a kegyeletes föláldozó
testvérnek, a szívére hallgató női léleknek; de épen az a cselekvése,
mely ilyenné emeli, tiszteletlennek, tökéletlennek mutatja hazája
iránt való viszonyában. Macbeth tökéletes mint hős, győzhetetlen az
ellenséggel szemben; de képtelen legyőzni saját szenvedélyeit,
tökéletlen mint uralkodó. Romeo és Julia tökéletesek mint
szerelmesek; de minden egyébben vakok s az élet komoly
föladataira nem valók. Toldi derék lélek és nagy bajvívó; de erős
egyénisége s rakonczátlan kedve annál kevésbbé fér meg az udvari
élet formáiban, minél újabbak lesznek ezek; Estéjének győzelmes
fölmagasodása a párbajban s féktelen dühe a dévajkodó apródok
szobájában: szomszédul mutatják kiválóságát és gyarlóságát. A
tragikai hősök bizonyos irányú tökéletességök daczára, vagy épen
ennek következtében, nem bírják magukban azt az öszhangot,
melylyel a világ közrendjébe sorakozhatnának. Tökéletlenségöknél
fogva vagy nem gondolnak vele, vagy fölötte érzik magukat;
voltaképen pedig csak ellenében állnak. Ezzel készen van a
szakítás, melyből a küzdelem és bukásuk támad. Ők a rend ellen; a
rendbe sorakozó összeség, az egyetemes, pedig ő ellenök.
Íme a tragikum két eleme: az egyetemes és az egyén, s az
utóbbinak két oldala: a tökéletes és a tökéletlen, más szóval: a kiváló
és a gyarló. Lássuk mindenekelőtt azt a mozzanatot, mely a tragikai
jelenségek iránt érdeklődésünket megragadja: az egyén kiválóságát.
II.
Az egyén kiválóságának széptani

formái.
Eszthetikai szempontból vizsgálódván, az egyesnek kiválóságát

természetesen a szépnek vagy egyik vagy másik fő formájában kell
keresnünk. Ezekben azt a tökéletességet, mely lelkünket vonzza; azt
a kiválóságot, mely képzeletünket meghatja; szóval azt a
szépségbeli mozzanatot, mely érdeklődésünket a tragikai alakok és
események iránt fölébreszti. A szépnek ez a két fő formája: a kellem
és a fenség. E két fogalom valamelyike alá vonható a természetnek
és művészetnek mindaz a jelensége, melyet a szépnek nevezete
megillet. A mi csak szép, az vagy kellemes, vagy fenséges voltánál
fogva az. A minek kiválósága gyönyörködtetve hat ránk, az csak
kellembeli, vagy fenségbeli kiválóság lehet.
E széptani két fő fogalom természetét és területét az a viszony
jelöli ki és határozza meg, mely minden szép dolognak két alkotó
eleme, az eszme és alak, között, különböző jelenségekben
különböző módon, mutatkozik. Attól a ponttól, a hol a merev alakba
életre ébresztőn beköszönt a lélek, egész addig, a hol a kettő között
teljes harmonia mutatkozik: a kellem országa terűl el. Innen kezdődik
a fenség területe, melyben az eszme uralkodóvá emelkedik az alak
fölött, s terjed egészen odáig, a hol nyilvánvalóvá lesz, hogy a kiváló
túlsúlyra jutott eszmét külső kép, alak többé kifejezni nem képes. A
szépség birodalma kezdődik ott, a hol eszme és alak találkoznak, s
véget ér ott, a hol teljesen elszakadnak egymástól. E nagy birodalom
egyik szélétől a másikig az eszme jelentősége szakadatlanúl
emelkedik s ezzel viszonya az alakhoz folyton folyvást módosúl.
Kettejök teljes egyensúlya, zavartalan öszhangja a közepén
mutatkozik s itt van a két területnek, a kellemnek és fenségnek,
határa.
A mint az alak merev lekötöttségéből az eszme ihletésére
fölenged; a mint az élet mozdulatai kifejezni kezdik a lelket: előáll a
kellem tüneménye. Midőn e széptani fogalmat Schelling a
művészeti fejlődés egy fokozatának jellemzésére alkalmazza, éles
szellemmel világosítja meg együtt mindakettőt. «A teremtő szellem
kezdetben a formába veszve jelenik meg, hozzáférhetetlenűl
elzárkózva s még nagyjában is durván. Minél inkább sikerűl neki a
maga egész teljességét egy alkotásban egyesíteni, annál jobban
enged hovatovább szigorából, s midőn az alakot tökéletesen
áthatotta, hogy elégűlten nyugszik benne, fölvidúl és gyöngéd
vonalakban mozogni kezd. Ez a legszebb virágzás és teljesség
állapota, mikor a tiszta edény tökéletes voltában áll előttünk, a
természeti szellem megszabadúl kötelékeiből s érzi rokonságát a a
lélekkel. Mintegy szelíd hajnali fénynyel, mely az egész alakot
megszállja, jelentkezik az érkező lélek: még nincs itt, de gyöngéd
mozdulatok enyhe játékaival minden készűl már elfogadására. A
merev körvonalak olvadoznak és lágyúlnak. Valami kedves valóság,
mely sem szellemi, sem érzéki, hanem megfoghatatlan, terjed el az
alakon, simúl hozzá körvonalaihoz, tagjainak minden hajlásához. Ez
a meg nem fogható s mégis mindenki által érzett valóság az, a mit a
görögök nyelve charisnak, a mienk kellemnek nevez.» E néhány sor
ép oly költőileg, mint élesen utal a kellem valójára. A külső mintegy
önkénytelen tolmácsává lesz a léleknek s készen és szivesen simúl
hozzá. Készsége szabadságot és rendet egynek mutat; a kényszer s
az ebből folyó meghasonlás száműzve vannak köréből. Hódolata
önkénytelennek, hódítása öntudatlannak látszik. A hol még nem
teljes is a harmonia eszme és alak közt, legalább a megvalósítására
való törekvés mindig jelen van. Épen e törekvésben rejlik a kellem
bátorító, üdítő, vidámító hatása, mely magyarázatát szolgáltatja a
plátói mondásnak, hogy: szeretet az, a mivel a szépet felfogjuk. Mert
bármi különbözők a kellem jelenségei s az általok keltett hatások:
ezek mindegyikének alapszinét a szeretet szolgáltatja. E
mozzanatok jelölik ki nagyjából a kellem területét és határait. Itt
találjuk a feslő rózsabimbót, a csörgő patakot, a vadgalamb búgását
ép úgy, mint Beranger és Petőfi dalait; Czuczor falusi kis leányát
karöltve Vörösmarty Szép Ilonkájával; Boëthos játszó gyermekét
s a vatikáni Ámort a medicei Vénus mellett.
A fenség képei már az eszmét és alakot megváltozott viszonyban
tüntetik föl. A kettő valódi egyensúlyának határát elhagyjuk. Az
eszme túlemelkedik az alakon, felsőségre jut fölötte s viszonyuk
bomlásnak indúl. A magamagának határt, a harmonia határát, vető
fejlődés mögöttünk marad. A fenség föléje látszik emelkedni az
érzéki világ törvényeinek, maga alatt hagyni mértékeit. Végetlen
eszme véges képben: megmérhetetlen nagyság s legyőzhetetlen
erő. Ki számítja ki a vihar erejét, ki tudja a tenger határát? Azonban
a végetlennek e véges nyilatkozásában kétségtelen ellenmondás
rejlik: a harmonia tárgyi igazsága megszűnik. Ez az ellenmondás
csak szubjektiv megoldást nyer a fenséges dolgok hatása által
képzeletünkre. Midőn szemben állunk velök, a kép végessége,
korlátolt volta eltűnni látszik; az alakot képzeletünk a végetlenig
tágítja ki, hogy alkalmasnak láthassa az eszme kifejezésére. Így áll
helyre s így marad fenn, a meddig maradhat, a harmonia.
Krisztusnak mindeneket magához ölelő szeretete, meghatott
képzeletünkben földi alakját is az isteni mindenhatóság színébe
öltözteti, mely az egész bűnös emberiséget megváltani képes volt.
III. Rikhárd ádáz szenvedélye úgy hat ránk, mintha képes lenne az
egész gyűlölt világot kiírtani; ha idejekorán, mindjárt kezdetben
fölmerűlne bennünk ereje végességének tudata s ebben bukásának
bizonyossága: a csodálat és félelem aligha venne erőt azon a
megnyugváson, melylyel a vele szembeszálló s őt előbb-utóbb
megsemmisítő hatalom megjelenését várnók. Midőn a harczos
tábort fölidézi a költő képzelete s olvasói elé varázsolja, «vassá
válnak az emberek, a sok dárda Bakonynyá»: ime a fantáziának
szinte öntudatlan munkája a fenség tényezőinek, az erőnek és
nagyságnak, túlzásában. A természet némely tüneménye segíti
képzeletünket s egyszersmind durván bár, de érthetőn példázza azt
a jelenséget, melyet a fenség elméletére nézve kiváló fontosnak
tartunk. A tengerre ereszkedő köd elmossa a kép éles határvonalát,
s a roppant víztömeget mintegy az ég végtelenségével olvasztja
egybe. Az alpesek tetejét borító hó fénye az égboltozat
világosságával játszik össze s az ormokon ülő felhő eltakarja és
képzeletünk előtt mérhetetlen magasságba emeli a csúcsot. A
mennydörgés hallatára úgy érezzük, mintha az egész mindenség
fölött tombolna, a mint távolodó zúgással elmorajlik; holott a
mindenségnek mily parányi részecskéje földünk s még ennek is mily
csekély területe az, melyet a zúgó felhők beárnyékolnak. Uralkodjék
bár a fenség képeiben akár a nagyság, akár az erő, hatalmas
benyomásaikra elvesztjük szokott mértékünket: a képpel összeolvad
az eszme, melynek a képzeletünk által kiterjesztett alak méltó
viselőjeűl tűnik föl.
Ilyen a fenség hatása, mely végességünk, törpeségünk,
gyöngeségünk éreztetésén át emel föl bennünket a végetlen
gondolatához. Mikor e gondolatot lelkünkbe fogadjuk, bizonyságát
érezzük, hogy részesei vagyunk. A mit a szellem elgondolni és
elképzelni tud: ahhoz fölemelkedett; a mi eltölti, az az övé. Fordítsuk
el némileg Vörösmarty szavait: «Sajátunknak mondhatjuk, a mit a
szív felfoghat magába.» Testi törpeségünk és gyöngeségünk
érzetéről lelki nagyságunk és erőnk tudatára emelkedünk. «Én is
festő vagyok»! tört ki Correggio a sixtina madonna előtt. Ebben a
végetlenhez való, csodálattal és gyönyörrel teljes fölemelkedésben
tetőzik és határozódik a fenség érzete. Területe a mondottakból
kimérhető. Véghatára ott van, a hol eszme és kép elszakadnak: az
alakot a képzelet szárnya sem emelheti föl többé az eszméhez.
Ez, rövid vázlatban, a kellem és fenség viszonya. Ott az alakba
köszöntő s azt elevenítő eszme, itt a képből kibontakozni készülő; ott
a valósúló harmonia, itt a fenyegetett. A kellem mindig törekvést
fejez ki a harmonia megalkotására, a fenség pedig feloldására. Az
szeretetet ébreszt, mikor tetszik; ez csodálatot és elragadtatást,
mikor fölemel.
Ez a szépnek két fő formája. E két formában nyilatkozhatik az a
kiválóság, az a tökéletesség, melyet a tragikum egyik
momentumának jelöltünk meg.
III.
A kellembeli kiválóság.
A szépnek az a formája, melyet kellemnek neveztünk, képezhet-

e bizonyos körülmények között tragikus mozzanatot? Lehet-e az a
kiválóság, mely a tragikai jelenségeknek egyik jellemző oldala,
kellembeli tökéletesség? Röviden: átmehet-e a kellemes a
tragikusba? Első pillanatra úgy látszik, mintha ez egyáltalában
lehetetlen lenne. Mikép indíthatna bárminemű meghasonlást az, a
minek fő jellemvonása épen a harmonia; ennek megvalósítása,
feltüntetése, éreztetése a czélja? A kellem épen nyugalmában és
elégűltségében mutatja az alakot átható eszmét. A szép, mely
ekként nyilatkozik, egészen magában elzárt, mitsem követelő, a lét
megnyugvását kifejező, magával és mindennel öszhangban levő. A
harmonia jelenségéből hogyan fejlődhetik már öszhangtalanság?
Épen az, a mi magának mértéket, határt tud szabni s teszi ezt nem
kényszerből, hanem jószántából és szabadsága szerint, hogyan
keveredhetik ellentétbe az egyetemessel?
Két úton is: két, természetében rejlő momentumnak
kiemelkedése által.
Az egyik épen egyessége, önelégűltsége, elzártsága. Ez a
mozzanat, a természetes fejlődés nyomán, oda vezet, a hol már
szemben találjuk az egyetemessel. Épen magával való
öszhangjának teljessége lesz meghasonlásának forrásává az
általánossal, mely külön voltának némi feladását követeli tőle. Ily
helyzetében a kellemest nem az alakjában nyilatkozó mérséklet
oldaláról tekintjük, hanem egyességre, önállóságra való
törekvésében. A nyugalom momentumát a törekvésé váltja fel, s az
érzés, melyet az így mutatkozó kellem ébreszt, megmarad ugyan
még mindig a szeretet alapján, de a zavartalan tetszésből és
örömből a nyugtalanabb féltésbe és szánalomba suhan át. Mert az
egyesnek ilyetén önelhatárolásában már veszélyes mértéktelenség
van; Zeising az isteni bitorlásának mondja. Önelégűltség nyilatkozik
benne, mely csak a mindenséget illeti meg. Ha nem támad is föl az
egyetemes ellen s nem száll is tudatosan szembe vele, de kevésbe
veszi, s a nélkül, hogy jogait és erejét ismerné, maga ellen ingerli. Ez
okból lehetetlen megmaradnia; ezen az úton siet vesztébe a szép.
Épen harmoniája s a magában való öntudatlan bizakodás, melyre a
harmonia vezette s a mely az egyetemes szempontjából
elbizottságnak tűnik föl, lesz veszedelmévé.
Alig idézhetnénk találóbb példát e jelenség magyarázatára
Melindánál. Naiv bizalommal, tapasztalatlan jósággal jelenik meg ifjú
szépségében az udvarnál, hol bűnös ármányok, léha és zabolátlan
erkölcsök uralkodnak s bősz szenvedélyek csatája készűl. Ebben
meg is törik s menthetetlenűl elhervad. Miért? Azt gondolta, hogy
egyszerű jóságában, nemes szeretetével élhet külön, magára; hogy
tiszta kelleme érintetlen maradhat attól a körtől, a Gertrud udvarától,
melybe kerűlt. Ezt az udvart nem ismeri; sőt szükségét sem érzi,
hogy igazán megismerje. Ártatlan bizalma nyilatkozik, midőn Ottóval
szemben, ki üldöztetését panaszolja, könnyekig lágyúl, s a
leghevesebb ostrom közben is, midőn a visszautasított herczeg
megátkozott jövendőjét emlegeti, nem tudja elnyomni s elhallgatni
szánalmát iránta. E könnyek, e szánalom egy belső öszhang
nyilatkozata, egy gyöngéd, bizalomteljes ártatlanságé, melyben
azonban Melinda romlása foganik meg. Úgy hiszi, vagy inkább érzi,
hogy tisztasága elég fegyver az ő külön léte határainak
megvédésére, holott a világban, melybe sorsa vetette, másféle
eszközökre lenne szüksége, azokhoz mértekre, melyek ott
használatban vannak: belátásra, erélyre, udvari módra.
Csalódásából foly tragikus sorsa. De naiv hite, tapasztalatlansága,
bízó törekvése, hogy olyan maradjon a milyen: mindez, fiatalsága és
szépsége mellett, kellemének alkatrésze, mely Ottó szenvedélyét
táplálja, s épen mint ilyen, bukásának forrása. Ez az okozati
összefüggés Melinda kelleme és bukása között, melyre Gyulai Pál
is rámutatott, midőn tapasztalatlanságát majdnem tragikai
tévedésnek mondja. – Kemény Zsigmond Tarnóczi Sáráját mily
vonzóvá, bájossá teszi titkolt szerelme, rejtegetett eszményképe!
Szívbeli állandósága, mely reménytelenűl őrzi titkát, ez a kitartó
hűség, ez a félénk elzárkózás, ez a magánakvalóság, átkává és
romlásává lesz magának és mindazoknak, a kiket szeret s a kik
szeretik. Titka: szerelme Mikes János iránt, áthatja egész valóját,
annak részévé lesz. Ha megosztaná valakivel, ha kilépne
elzárkózottságának bűvös köréből, talán megmenekedhetnék; így
sem magát nem tudja megőrizni, sem szeretteit.
A kellemhez, mely magát egyességében, különösségében az
általánossal szemben érvényesíteni törekszik, közel áll egy másik
jelenség. Az, mikor a látszatra kellemes nem öntudatlan
természetességgel szigeteli el magát, hanem ellenkezőleg öntudatos
természetlenséggel; mikor egyességének követelése épen nem
önkénytelen. Törekvése abban fejezhető ki, hogy nem enged helyet
annak az eszmének, annak a tartalomnak, melyet természet szerint
befogadnia s képviselnie kell. Távol tartja magától s ebben áll
különössége, szembeszállása a világgal, ellentéte a természeti
szükségszerűséggel. Ez az ellentét amott öntudatlan volt, itt
öntudatos. Gondoljunk a nőre, a ki nem szeret, mert nem akar
szeretni. De az ilyen alakban az öszhang csak látszólagos, melyet a
hidegség, nyugalom, biztosság igyekeznek elhitetni velünk. A kellem
épen ennélfogva elenyészően csekély. A dolog természetéből foly,
hogy előbb-utóbb helyt kell adnia lelke természetes tartalmának. A
körülmények kényszerítik, hogy magába fogadja az eszmét, melynek
képviselésére alkotva van. Látni való, hogy e fejlődés egészen
ellenkező irányú a tragikummal: az összeütközésből, melybe az
egyetemessel keveredett, nem megsemmisűlésére, hanem belső
harmoniájának megvalósítására vezet. Rákosi Jenő Stellájával egy
ilyetén ellenmondásnak tragikai irányú megoldására látszott
törekedni; de kénytelen volt félútról visszafordúlni s az egész,
borzalomig csigázott cselekvényt álomnak tüntetve föl,
kiengesztelően végezni meséjét. Mikor ugyanis a fejtegetett alakok
végre befogadják természetes tartalmukat, csak akkor lesznek
igazán kellemesekké; de ugyanakkor magukkal ellenmondásba
keveredvén, többnyire komikusakká. Példát Lope de Vega Donna
Juanája a Megvetés csodájában, azután Molière Eliszi herczegnője
s Moreto Donna Dianája szolgáltathat.
A kellem azonban még egy más esetben is átcsaphat a
tragikumba: teljes virága, a báj, édes illatával bódítóvá és
vészthozóvá lehet. A valójában rejlő erőbeli elem felülkerekedhetik s
megronthatja harmoniáját. Kifejezésre jut a hódítás vágyában, a rá
való törekvésben. De hát nem jogosúlt-e a szépnek ez a törekvése?
Semmi esetre sem akkor, ha legfőbb hatalom gyanánt kivánván
mutatkozni, felforgatással fenyegeti a természeti rendet, melyben
nem ő van hivatva döntő szerepre, s az erkölcsit, melyet önzésénél
fogva alárendeltnek tekint és nem tart tiszteletben. Szóval a báj
tragikussá lesz, mikor mintegy önmagából kiszállva csábítóvá válik:
a csábítás ellenállhatatlan erejével nyilatkozó báj. Az «arany
pohárban nyújtott méregital». A szirének, kik csábító dalukkal
kikerűlhetetlen vészbe csalták a hajósokat, az antik képzeletben
halálgéniuszokká alakúltak; a madárszárnyú és madárlábú
nőalakokkal gyakran találkozunk görög és etrusk síremlékeken.
A költészetben ott van Shakspere Kleopatrája. A csábító bájnak
teljes fegyverzetével, erőszakos czélzatával lép föl mindjárt a
búcsúzás jelenetében a világhódító Antonius, e római Hercules,
ellen. Báját, igéző hatalmát igazi asszony voltának köszöni; egyesűl
benne szépség és szellem, izzó és gyorsan változó érzések
ellenmondásai, keleti hazájának bujasága s korának mesterkélt
mívelődése, királyi büszkeség és léha ingatagság, mohó érzékiség
és forró ragaszkodás. Ez arany-szálak hálójába kerül a hős
Antonius, ki «kardjának hegyén bírá a földet». Kettejök viszonyát
mindjárt a dráma legelső szavaiban következőkép jellemzi Philo:
Nem! a vezér ez őrjöngő szerelme

Túlmegy határin; ama hős szemek,
Mik csatarend és harczi zaj fölött
Lángoltak, érczet öltött Mars gyanánt,
Most szolgalánggal oltároznak egy
Napsütte arcz előtt; a hős kebel,
Mely nagy csaták küzdelmi közt feszíté
Pánczéla kapcsát szerte, meglohadt most,
És legyező- s fuvóvá aljasúlt:
Lehűtni egy czigánynő lánghevét!
Természetes, hogy Antoniusnak, a ki ekként megtagadta magát,

vesznie kell. Kleopatra teszi tönkre: együtt az erkölcsi és hatalmi
elvet, melynek hordozására született, családi életét és uraságát,
magán és közhivatását. Midőn ily rendkívüli erőt ront meg bájával,
ez jogosulatlan hatalommá lesz, melynek szintén buknia kell.
Asszonyi lényéből, melylyel Antoniust behálózta s a mely oly
ellenállhatatlanná teszi, foly árulása Actiumnál, hol az ő hajóinak
Octaviushoz pártolása dönti el a csata sorsát. Antonius vad átkokkal
halmozza el s halállal készűl büntetni hitlenségét. De varázsa még
ekkor is él szívén, mert öngyilkosságának költött hírére azonnal
kiengesztelődik iránta s elvesztésén kétségbeesve dűl kardjába.
Schlegel ide vonatkozólag azt a megjegyzést teszi, hogy: «midőn
ők így egymásért halnak, megbocsátjuk nekik, hogy egymásért
éltek». Ez tévedés, mert csak Antonius volt az, a ki Kleopatráért halt
meg; kedvese, kinek csábító szerelme erkölcsi méltóság híján volt s
inkább érzékiségben mint szívbeli hajlamban gyökerezett, maga
miatt és magáért hal. Vesznie kell, midőn oly hatalommal ütközik
össze, melylyel szemben csábító bája tehetetlen. Ilyen pedig a
mérsékelt, okos, megfontolt, hideg Octavius. Rá nézve Kleopatra
nem a csábító asszony, csak a hadi fogoly, kinek a cæsar
diadalmenetét ékesíteni vagy halnia kell. Antonius az «agg Nil
kigyójának» szokta nevezni kedvesét. Romlásának képe, a mi által
elveszett, egyszersmind életének képe: a kigyó. Ez a csábítás
legrégibb szimboluma; Michel Angelo freskója a Sixtina
mennyezetén, az első emberpár esete, a kigyó testét összeolvasztva
ábrázolja a női testtel. A kigyó lett a művészi képzeletben Kleopatra
alakjának is szinte elválhatatlan járuléka: Paolo Veronesetől és
Guido Renitől kezdve a mi országos képtárunk hollandi Bys
Jánosáig valamennyi művész oda képzelte a csábító mellre.
Antonius és Kleopatrát Shakspere már életének őszén írta; de
lángesze alig mutatkozik valahol frisebbnek és többoldalúnak, mint e
tragédiában. Egyetlen nőalakja sem tünteti föl a női természet egész

Download Advances In Machine Learning Deep Learning Based Technologies Selected Papers In Honour Of Professor Nikolaos G Bourbakis Vol 2 Learning And Analytics In Intelligent Systems 23 George A Tsihrintzis online ebook texxtbook full chapter pdf

Uploaded by

Copyright:

Available Formats

Download Advances In Machine Learning Deep Learning Based Technologies Selected Papers In Honour Of Professor Nikolaos G Bourbakis Vol 2 Learning And Analytics In Intelligent Systems 23 George A Tsihrintzis online ebook texxtbook full chapter pdf

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Download Advances In Machine Learning Deep Learning Based Technologies Selected Papers In Honour Of Professor Nikolaos G Bourbakis Vol 2 Learning And Analytics In Intelligent Systems 23 George A Tsihrintzis online ebook texxtbook full chapter pdf

Uploaded by

Copyright:

Available Formats

Advances in Machine Learning Deep

Learning based Technologies Selected

Advances in Artificial Intelligence based Technologies

Machine Learning Techniques and Analytics for Cloud

Machine Learning and Deep Learning Techniques in

Deep Learning in Cancer Diagnostics: A Feature-based

Design of Intelligent Applications using Machine

Advances in Deep Learning, Artificial Intelligence and

Recent Advances in Internet of Things and Machine

Computer Vision and Machine Learning in Agriculture

Learning and Analytics in Intelligent

The main aim of the series is to make available a publication of books in

Advances in Machine Learning/Deep

ISSN 2662-3447 e-ISSN 2662-3455

© The Editor(s) (if applicable) and The Author(s), under exclusive

The use of general descriptive names, registered names, trademarks,

This Springer imprint is published by the registered company Springer

In 1959, Arthur Samuel [1, 2] introduced the term Machine Learning

3. J.E. Ormrod, Human Learning, Pearson, 8th edition, ISBN-13: 978-

4. M. Nielsen, Neural Nets and Deep Learning (2019), http://​

5. K. D. Foote, A Brief History of Machine Learning (2019) (https://​

6. History of Machine Learning, https://​www.​doc.​ic.​ac.​uk/​~jce317/​

7. Introduction to Neural Nets, https://​www.​doc.​ic.​ac.​uk/​~jce317/​

8. B. Grossfeld, Deep learning vs machine learning: a simple way to

9. G. A. Tsihrintzis and L. C. Jain (Eds.), Machine Learning Paradigms

10. N. Bourbakis, Artificial Intelligence (AI) and its Impact to

12. Chandrasekar Vuppalapati, Democratization of Artificial

1. Introduction to Advances in Machine

George A. Tsihrintzis (Corresponding author)

1.1 Editorial Note

II. Machine Learning/Deep Learning in Education

III. Machine Learning/Deep Learning in Security

IV. Machine Learning/Deep Learning in Time Series Forecasting

V. Machine Learning in Video Coding and Information Extraction

The book at hand aims at updating the relevant research

2. Semi-supervised Feature Selection Method for

Ferdinando Di Martino (Corresponding author)

2.2 The FS-EFCM Algorithm

Fig. 2.1 Logical overview of the algorithm

Listing 1. Algorithm: FS-EFCM

2.2.2 Initial Parameter Setting

Listing 2. Algorithm: SelectDeltaThreshold

2.3 Experimental Results

2.3.2 Feature Selection

Abomin Absorb Accept Ach Admir Affect Affection Afflict Afraid

Number of Threshold Features Relevant features New Feature Features

Table 2.3 Results obtained launching FS-EFCM

Cycle Number of Number of Feature Features Stop iteration

Class Documents belonging to Cluster Documents assigned to the Well classified

Table 2.6 Classification performance indices

Measure Value (%)

Class Documents belonging to Cluster Documents assigned to the Well classified

Measure Value (%)

Author: Zsolt Beöthy

Release date: July 29, 2022 [eBook #68633]

Original publication: Hungary: Kisfaludy-Társaság, 1885

Credits: Albert László from page images generously made available

*** START OF THE PROJECT GUTENBERG EBOOK A TRAGIKUM

Az életnek és művészetnek minden tragikai jelensége, végső

Az egyén kiválóságának széptani

4. M. Nielsen, Neural Nets and Deep Learning (2019), http://

5. K. D. Foote, A Brief History of Machine Learning (2019) (https://

6. History of Machine Learning, https://www.doc.ic.ac.uk/~jce317/

7. Introduction to Neural Nets, https://www.doc.ic.ac.uk/~jce317/