Lecture Notes in Computer Science 9124

Commenced Publication in 1973Founding and Former Series Editors:Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board

David HutchisonLancaster University, Lancaster, UK

Takeo KanadeCarnegie Mellon University, Pittsburgh, PA, USA

Josef KittlerUniversity of Surrey, Guildford, UK

Jon M. KleinbergCornell University, Ithaca, NY, USA

Friedemann MatternETH Zurich, Zürich, Switzerland

John C. MitchellStanford University, Stanford, CA, USA

Moni NaorWeizmann Institute of Science, Rehovot, Israel

C. Pandu RanganIndian Institute of Technology, Madras, India

Bernhard SteffenTU Dortmund University, Dortmund, Germany

Demetri TerzopoulosUniversity of California, Los Angeles, CA, USA

Doug TygarUniversity of California, Berkeley, CA, USA

Gerhard WeikumMax Planck Institute for Informatics, Saarbrücken, Germany

More information about this series at http://www.springer.com/series/7412

Marzena Kryszkiewicz • Sanghamitra BandyopadhyayHenryk Rybinski • Sankar K. Pal (Eds.)

Pattern Recognitionand Machine Intelligence6th International Conference, PReMI 2015Warsaw, Poland, June 30 – July 3, 2015Proceedings

123

EditorsMarzena KryszkiewiczInstitute of Computer ScienceWarsaw University of TechnologyWarsawPoland

Sanghamitra BandyopadhyayMachine Intelligence UnitIndian Statistical InstituteKolkata, West BengalIndia

Henryk RybinskiInstitute of Computer ScienceWarsaw University of TechnologyWarsawPoland

Sankar K. PalIndian Statistical InstituteKolkata, West BengalIndia

ISSN 0302-9743 ISSN 1611-3349 (electronic)Lecture Notes in Computer ScienceISBN 978-3-319-19940-5 ISBN 978-3-319-19941-2 (eBook)DOI 10.1007/978-3-319-19941-2

Library of Congress Control Number: 2015940415

LNCS Sublibrary: SL6 – Image Processing, Computer Vision, Pattern Recognition, and Graphics

Springer Cham Heidelberg New York Dordrecht London© Springer International Publishing Switzerland 2015This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of thematerial 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 informationstorage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology nowknown or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoes not imply, even in the absence of a specific statement, that such names are exempt from the relevantprotective 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 arebelieved to be true and accurate at the date of publication. Neither the publisher nor the authors or the editorsgive a warranty, express or implied, with respect to the material contained herein or for any errors oromissions that may have been made.

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media(www.springer.com)

Preface

This volume contains the papers selected for presentation at the 6th InternationalConference on Pattern Recognition and Machine Intelligence (PReMI 2015), whichwas held at Warsaw University of Technology, Warsaw, Poland, during June 30 – July3, 2015.

PReMI is a conference series that started in 2005. Held every two years, PReMIprovides an international forum for exchanging scientific, research, and technologicalachievements in pattern recognition, machine intelligence, and related fields. In par-ticular, major areas selected for PReMI 2015 include pattern recognition, machineintelligence, image processing, retrieval and tracking, data mining techniques for large-scale data, fuzzy computing, rough sets, bioinformatics, and applications of artificialintelligence. In addition, two special sessions were organized; namely, Special Sessionon Data Mining Techniques for Large-Scale Data and Special Session on Scalability ofRough Set Methods. Four plenary keynote talks and two tutorials were delivered. ThePReMI 2015 conference was accompanied by the Industrial Session on MachineIntelligence and Big Data in the Industry.

PReMI 2015 received 90 submissions that were carefully reviewed by three or moreProgram Committee members or external reviewers. Papers submitted to special ses-sions were subject to the same reviewing procedure as those submitted to regularsessions. After a rigorous reviewing process, 54 papers were accepted for presentationat the conference and publication in the PReMI 2015 proceedings volume. This volumealso contains one invited paper and three extended abstracts by the plenary keynotespeakers.

It is truly a pleasure to thank all those people who helped this volume to come intobeing and to turn PReMI 2015 into a successful and exciting event. In particular, wewould like to express our appreciation for the work of the PReMI 2015 ProgramCommittee members and external reviewers who helped assure the high standards ofaccepted papers. We would like to thank all the authors of PReMI 2015, without whosehigh-quality contributions it would not have been possible to organize the conference.We are grateful to the organizers of the PReMI 2015 special sessions: Julian Szymańskiand Marcin Błachnik (Special Session on Data Mining Techniques for Large-ScaleData) as well as Jarosław Stepaniuk (Special Session on Scalability of Rough SetMethods). We would also like to express our appreciation to the Organizing Committeechairs (Robert Bembenik and Łukasz Skonieczny) for their involvement in all theorganizational matters related to PReMI 2015 as well as the creation and maintenanceof the conference website. We are grateful to Bożenna Skalska and Joanna Konczak fortheir administrative work.

Special thanks go to Andrzej Skowron who was spiritus movens for this conferenceto take place in Warsaw and become a successful scientific event in Warsaw. Fur-thermore, we want to thank the Industrial Session chairs (Piotr Gawrysiak and DominikRyżko), and we also gratefully acknowledge the generous help of the remaining PReMI

2015 chairs – Piotr Andruszkiewicz, Grzegorz Protaziuk, Santanu Chaudhury, SergeiKuznetsov, as well as of the Steering Committee members – Malay K. Kundu andAndrzej Skowron. We wish to express our thanks to George Karypis, Sankar K. Pal,Roman Słowiński, and Xin Yao for accepting to be plenary speakers at PReMI 2015.We also thank the PReMI 2015 tutorial speakers – Gerald Schaefer and SantanuChaudhury.

Our thanks are due to Alfred Hofmann of Springer for his continuous support and toAnna Kramer and Christine Reiss for their work on the proceedings.

We believe that the proceedings of PReMI 2015 will be a valuable source ofreference for your ongoing and future research activities.

June – July 2015 Marzena KryszkiewiczSanghamitra Bandyopadhyay

Henryk RybinskiSankar K. Pal

VI Preface

Organization

PReMI 2015 was organized by the Institute of Computer Science, Warsaw Universityof Technology, Warsaw, Poland, in cooperation with Machine Intelligence Unit, IndianStatistical Institute, Kolkata, India, and Center for Soft Computing Research, IndianStatistical Institute, Kolkata, India.

PReMI 2015 Conference Committee

Honorary Chair

Sankar K. Pal Indian Statistical Institute, Kolkata, India

Steering Committee

Malay K. Kundu Indian Statistical Institute, Kolkata, IndiaAndrzej Skowron University of Warsaw, Poland

General Chair

Henryk Rybinski Warsaw University of Technology, Poland

Program Chairs

SanghamitraBandyopadhyay

Indian Statistical Institute, Kolkata, India

Marzena Kryszkiewicz Warsaw University of Technology, Poland

Organizing Chairs

Robert Bembenik Warsaw University of Technology, PolandŁukasz Skonieczny Warsaw University of Technology, Poland

Special Session and Tutorial Chairs

Piotr Andruszkiewicz Warsaw University of Technology, PolandGrzegorz Protaziuk Warsaw University of Technology, Poland

Industrial Session Chairs

Piotr Gawrysiak Warsaw University of Technology, PolandDominik Ryżko Warsaw University of Technology, Poland

International Liaison Chairs

Santanu Chaudhury Indian Institute of Technology, Delhi, IndiaSergei O. Kuznetsov Higher School of Economics, Moscow, Russia

Program Committee

Cesare Alippi Politecnico di Milano, ItalyPiotr Andruszkiewicz Warsaw University of Technology, PolandAnnalisa Appice Università degli Studi di Bari, ItalyR. Venkatesh Babu Indian Institute of Science, Bangalore, IndiaSanghamitra

BandyopadhyayIndian Statistical Institute, Kolkata, India

Sivaji Bandyopadhyay Jadavpur University, Kolkata, IndiaMinakshi Banerjee RCC Institute of Information Technology,

Kolkata, IndiaJan G. Bazan University of Rzeszow, PolandRobert Bembenik Warsaw University of Technology, PolandSambhu Nath Biswas Indian Statistical Institute, Kolkata, IndiaMarcin Błachnik Silesian University of Technology, Katowice, PolandIlona Bluemke Warsaw University of Technology, PolandMichelangelo Ceci Università degli Studi di Bari, ItalyBasabi Chakraborty Iwate Prefectural University, JapanGoutam Chakraborty Iwate Prefectural University, JapanMihir Chakraborty CSCR, Indian Statistical Institute, Kolkata, IndiaAmitava Chatterjee Jadavpur University, Kolkata, IndiaSung-Bae Cho Yonsei University, KoreaAnanda Shankar

ChowdhuryJadavpur University, Kolkata, India

Paweł Cichosz Warsaw University of Technology, PolandDavide Ciucci University of Milano-Bicocca, ItalyAndrzej Czyżewski Gdansk University of Technology, PolandAmit Das Bengal Engineering and Science University, Shibpur,

IndiaPartha Pratim Das India Institute of Technology, Kharagpur, IndiaSukanta Das Bengal Engineering and Science University, Shibpur,

IndiaRajat De Indian Statistical Institute, Kolkata, IndiaAndries Engelbrecht University of Pretoria, South AfricaPaolo Gamba University of Pavia, ItalyTomasz Gambin Warsaw University of Technology, PolandNiloy Ganguly Indian Institute of Technology, Kharagpur, IndiaBernhard Ganter Dresden University of Technology, GermanyPiotr Gawkowski Warsaw University of Technology, PolandAshish Ghosh Indian Statistical Institute, Kolkata, IndiaKuntal Ghosh Indian Statistical Institute, Kolkata, IndiaSusmita Ghosh Jadavpur University, Kolkata, IndiaJarek Gryz York University, Toronto, CanadaJerzy Grzymała-Busse University of Kansas, Kansas, USAFrancisco Herrera University of Granada, Spain

VIII Organization

Andrzej JankowskiC.V. Jawahar International Institute of Information Technology,

Hyderabad, IndiaMario Koeppen Kyushu Institute of Technology, JapanJakub Koperwas Warsaw University of Technology, PolandJacek Koronacki Polish Academy of Sciences, PolandBożena Kostek Gdansk University of Technology, PolandStanisław Kozielski Silesian University of Technology, Gliwice, PolandKrzysztof Krawiec Poznań University of Technology, PolandMarzena Kryszkiewicz Warsaw University of Technology, PolandMalay Kundu Indian Statistical Institute, Kolkata, IndiaSergei O. Kuznetsov National Research University Higher School of

Economics, Moscow, RussiaHalina Kwaśnicka Wroclaw University of Technology, PolandWitold Łukaszewicz University of Warmia and Mazury in Olsztyn, PolandPradipta Maji Indian Statistical Institute, Kolkata, IndiaDonato Malerba Università degli Studi di Bari, ItalyDeba Prasad Mandal Indian Statistical Institute, Kolkata, IndiaMichał Marcińczuk Wroclaw University of Technology, PolandFrancesco Masulli Università di Genova, ItalyUjjwal Maulik Jadavpur University, Kolkata, IndiaJesús Medina-Moreno University of Cádiz, Puerto Real, SpainEvangelos Milios Dalhousie University, Halifax, CanadaPabitra Mitra Indian Institute of Technology, Kharagpur, IndiaSuman Mitra Dhirubhai Ambani Institute of Information and

Communication Technology, Gujarat, IndiaSushmita Mitra Indian Statistical Institute, Kolkata, IndiaMikołaj Morzy Poznań University of Technology, PolandTadeusz Morzy Poznań University of Technology, PolandHiroshi Motoda Osaka University, JapanDariusz Mrozek Silesian University of Technology, Gliwice, PolandJayanta Mukherjee Indian Institute of Technology, Kharagpur, IndiaMieczysław Muraszkiewicz Warsaw University of Technology, PolandC.A. Murthy Indian Statistical Institute, Kolkata, IndiaTomoharu Nakashima Osaka Prefecture University, JapanMirco Nanni KDD-Lab ISTI-CNR Pisa, ItalyAmedeo Napoli Loria, Vandoeuvre-lès-Nancy Cedex, FranceY. Narahari Indian Institute of Science, Bangalore, IndiaMita Nasipuri Jadavpur University, Kolkata, IndiaHung Son Nguyen Warsaw University, PolandRobert Nowak Warsaw University of Technology, PolandWłodzimierz Ogryczak Warsaw University of Technology, PolandMichał Okoniewski Functional Genomics Center Zurich, SwitzerlandAndrzej Pacut Warsaw University of Technology, PolandDebnath Pal Indian Institute of Science, Bangalore, IndiaGabriella Pasi University of Milano-Bicocca, Italy

Organization IX

James F. Peters University of Manitoba, Winnipeg, CanadaAlfredo Petrosino University of Naples Parthenope, ItalyGrzegorz Protaziuk Warsaw University of Technology, PolandVijay V. Raghavan NSF I/UCRC, USASheela Ramanna University of Manitoba, Winnipeg, CanadaZbigniew Raś University of North Carolina, Charlotte, USA

and Warsaw University of Technology, PolandSubhra Sankar Ray Indian Statistical Institute, Kolkata, IndiaJadwiga Rogowska McLean Hospital/Harvard Medical School, Belmont,

USAPrzemysław Rokita Warsaw University of Technology, PolandHenryk Rybinski Warsaw University of Technology, PolandPunam Saha University of Iowa, USADebashis Sen National University of Singapore, SingaporePatrick Siarry Université Paris-Est Créteil, FranceJaya Sil Bengal Engineering and Science University, Shibpur,

IndiaWładyslaw Skarbek Warsaw University of Technology, PolandŁukasz Skonieczny Warsaw University of Technology, PolandAndrzej Skowron Warsaw University, Warsaw, PolandDominik Ślęzak University of Warsaw, Poland and Infobright Inc.,

Warsaw, PolandRoman Słowiński Poznań University of Technology, PolandJarosław Stepaniuk Bialystok University of Technology, PolandPonnuthurai Suganthan National Technological University, SingaporeZbigniew Suraj University of Rzeszow, PolandShamik Sural Indian Institute of Technology, Kharagpur, IndiaAndrzej Szałas Warsaw University, PolandMarcin Szczuka Warsaw University, PolandStan Szpakowicz University of Ottawa, CanadaJulian Szymański Gdansk University of Technology, PolandB. Uma Shankar Indian Statistical Institute, Kolkata, IndiaBrijesh Verma CQ University, AustraliaSlawomir Zadrożny Polish Academy of Sciences, Warsaw, PolandWlodek Zadrozny University of North Carolina, Charlotte, USAMaciej Zakrzewicz Poznań University of Technology, PolandWojciech Ziarko University of Regina, Canada

X Organization

External Reviewers

Alcalde, CristinaArtiemjew, PiotrBanerjee, AbhirupBanerjee, SanghitaBenítez Caballero, María JoséBetliński, PawełChu, HenryDe, ArijitDembski, JerzyDemidova, ElenaDiaz, ElizabethGamba, PaoloGambin, TomaszGarcía-Osorio, CésarGawkowski, PiotrGuerra, FrancescoKożuszek, Rajmund

Markkassery, SreejithMartincic-Ipsic, SandaMeina, MichalMondal, AjoyNayak, LosianaPal, Jayanta KumarPatra, Braja GopalPaul, SushmitaPio, GianvitoRoy, RahulRoy, ShaswatiRupino Da Cunha, PauloTerziyan, VaganTrillo Lado, RaquelTrzciński, TomaszŚwieboda, Wojciech

Organization XI

Invited Talks

Granular Mining and Rough-Fuzzy Computing:Data to Knowledge and Big Data Issues

Sankar K. Pal

Machine Intelligence Unitand

Center for Soft Computing ResearchIndian Statistical InstituteKolkata 700108, India

http://www.isical.ac.in/sankar

Extended Abstract

Pattern recognition and data mining in the framework of machine intelligence areexplained. The role of rough sets in uncertainty handling and granular computing ishighlighted. Relevance of its integration with fuzzy sets to result in a stronger paradigmfor uncertainty handling is explained. Generalized rough sets, rough-fuzzy entropy,different f-information measures, and fuzzy granular social network (FGSN) model aredescribed. FGSN handles the uncertainty arising from vaguely defined closeness orrelations of the actors (nodes). Various measures towards this are stated.

Rough-fuzzy image entropy takes care of the fuzziness in boundary regions as well asthe rough resemblance among nearby pixels and gray levels. Rough-fuzzy case generationwith variable reduced dimension is useful for mining data sets with large dimension andsize. Fuzzy granular model of social networks provides a generic platform for its analysis.Fuzzy-rough communities, detected thereby, are more significant when the degree ofoverlapping between communities increases. f-information measures quantify well themutual information in efficient feature selection, and the conditional information inmeasuring the goodness of community structures in network mining. These characteristicsare demonstrated for tasks like video tracking, social network analysis and gene/microRNA selection. The role of different kinds of granules is illustrated as well as theconcepts of fuzzy granular computing and granular fuzzy computing.

The talk concludes mentioning their relevance in handling Big data, the challengingissues and the future directions of research.

References

1. Pal, S.K., Mitra, P.: Case generation using rough sets with fuzzy representation. IEEE Trans.Knowl. Data Eng. 16(3), 292–300 (2004)

2. Sen, D., Pal, S.K.: Generalized rough sets, entropy and image ambiguity measures. IEEETrans. Syst. Man Cyberns. Part B 39(1), 117–128 (2009)

3. Maji, P., Pal, S.K.: Feature selection using f-Information measures in fuzzy approximationspaces. IEEE Trans. Knowl. Data Eng. 22(6), 854–867 (2010)

4. Pal, S.K., Meher, S.K., Dutta, S.: Class-dependent rough-fuzzy granular space, dispersionindex and classification. Pattern Recogn. 45(7), 2690–2707 (2012)

5. Pal, S.K.: Granular mining and rough-fuzzy pattern recognition: a way to naturalcomputation (feature article). IEEE Intell. Inform. Bull. 13(1), 3–13 (2012)

6. Kundu, S., Pal, S.K.: FGSN: fuzzy granular social networks - model and applications. Inf.Sci. 314, 100–117 (2015). doi:10.1016/j.ins.2015.03.065

7. Kundu, S., Pal, S.K.: Fuzzy-rough community in social networks. Pattern Recogn. Lett.doi:10.1016/j.patrec.2015.02.005

XVI S.K. Pal

Constructive Learning of Preferenceswith Robust Ordinal Regression

Roman Słowiński

Institute of Computing Science, Poznań University of Technology, 60-965 Poznań,and Systems Research Institute, Polish Academy of Sciences,

01-447 Warsaw, Polandroman.slowinski@cs.put.poznan.pl

Extended Abstract

The talk is devoted to preference learning in Multiple Criteria Decision Aiding. It iswell known that the dominance relation established in the set of alternatives (also calledactions, objects, solutions) is the only objective information that comes from aformulation of a multiple criteria decision problem (ordinal classification, or ranking, orchoice, with multiobjective optimization being a particular instance). While dominancerelation permits to eliminate many irrelevant (i.e., dominated) alternatives, it does notcompare completely all of them, resulting in a situation where many alternatives remainincomparable. This situation may be addressed by taking into account preferences of aDecision Maker (DM). Therefore, all decision-aiding methods require some preferenceinformation elicited from a DM or a group of DMs. This information is used to buildmore or less explicit preference model, which is then applied on a non-dominated set ofalternatives to arrive at a recommendation (assignment of alternatives to decisionclasses, or ranking of alternatives from the best to the worst, or the best choice)presented to the DM. In practical decision aiding, the process composed of preferenceelicitation, preference modeling, and DM’s analysis of a recommendation, loops untilthe DM accepts the recommendation or decides to change the problem setting. Such aninteractive process is called constructive preference learning.

I will focus on processing DM’s preference information concerning multiplecriteria ranking and choice problems. This information has the form of pairwisecomparisons of selected alternatives. Research indicates that such preference elicitationrequires less cognitive effort from the DM than direct assessment of preference modelparameters (like criteria weights or trade-offs between conflicting criteria). I willdescribe how to construct from this input information a preference model thatreconstructs the pairwise comparisons provided by the DM. In general, construction ofsuch a model follows logical induction, typical for learning from examples in AI. Incase of utility function preference models, this induction translates into ordinalregression. I will show inductive construction techniques for two kinds of preferencemodels: a set of utility (value) functions, and a set of “if. . ., then. . .” monotonicdecision rules. An important feature of these construction techniques is identification ofall instances of the preference model that are compatible with the input preferenceinformation – this permits to draw robust conclusions regarding DM’s preferences

when any of these models is applied on the considered set of alternatives. Thesetechniques are called Robust Ordinal Regression and Dominance-based Rough SetApproach.

I will also show how these induction techniques, and their corresponding models,can be embedded into an interactive procedure of multiobjective optimization,particularly, in Evolutionary Multiobjective Optimization (EMO), guiding the searchtowards the most preferred region of the Pareto-front.

References

1. Branke, J., Greco, S., Słowiński, R., Zielniewicz, P.: Learning value functions in interactiveevolutionary multiobjective optimization. IEEE Trans. Evol. Comput. 19(1), 88–102 (2015)

2. Corrente, S., Greco, S., Kadziński, M., Słowiński, R.: Robust ordinal regression in preferencelearning and ranking. Mach. Learn. 93, 381–422 (2013)

3. Figueira, J., Greco, S., Słowiński, R.: Building a set of additive value functions representinga reference preorder and intensities of preference: GRIP method. Eur. J. Oper. Res. 195, 460–486 (2009)

4. Szeląg, M., Greco, S., Słowiński, R.: Variable consistency dominance-based rough setapproach to preference learning in multicriteria ranking. Inf. Sci. 277, 525–552 (2014)

5. Słowiński, R., Greco, S., Matarazzo, B.: Rough-set-based decision support. In: Burke, E.K.,Kendall, G. (eds.) Search Methodologies: Introductory Tutorials in Optimization andDecision Support Techniques, Chap. 19, 2nd edn., pp. 557–609. Springer, New York (2014)

XVIII R. Słowiński

Ensemble Approaches in Learning

Xin Yao

CERCIA, School of Computer ScienceUniversity of Birmingham, Birmingham B15 2TT, UK

http://www.cs.bham.ac.uk/~xin

Extended Abstract

Designing a monolithic system for a large and complex learning task is hard. Divide-and-conquer is a common strategy in tackling such large and complex problems [1,2].Ensembles can be regarded an automatic approach towards automatic divide-and-conquer [3,4]. Many ensemble methods, including boosting [5], bagging [6], negativecorrelation [4], etc., have been used in machine learning and data mining for manyyears. This talk will describe three research topics in ensemble learning, i.e., multi-objective learning [7,8], online learning with concept drift [9,10], and multi-classimbalance learning [11,12]. Given the important role of diversity in ensemble methods[13,14], some discussions and analyses will be given to gain a better understanding ofhow and when diversity may help ensemble learning.

Multi-objective learning might first sound strange because the sole objective oflearning should be to maximise the generalisation ability of learned models. However,in ensemble learning, we are interested in finding an ensemble of accurate and diverseindividual learner. The accuracy and diversity naturally become two objectives. Whileone could aggregate these two objectives into a single function using a weighted sum, itis often very difficult in practice to tune the weights appropriately. An alternative toaggregating two objectives into one is to use a multi-objective optimisation algorithmto learn the two objectives simultaneously [7]. This is advantageous because a multi-objective optimisation algorithm will find a set of non-dominated solutions, rather thanjust a single solution, which can naturally be used as individual learners in anensemble. In a sense, multi-objective ensemble learning natural solves the problem ofdetermining what individuals should be in an ensemble. Once we have the multi-objective learning framework, it is very straightforward to add additional objectives,e.g., by adding an additional regularisation objective [8].

In the real world, data streams are very difficult to deal with, especially theirunderlying data distributions change, e.g., with concept drifts. What was correct longtime ago may not be correct anymore, and what was incorrect might become correctnow. It is always a huge challenge to deal with such concept drifts in online learning ofdata streams. Ensemble methods have long been used in online learning with conceptdrifts. However, there were few analyses of different roles that the diversity playsduring online learning. A detailed analysis of its roles [9] can actually provide insightinto different roles of the diversity at different stages of online learning. Such insight

can be exploited to develop new online ensemble learning algorithms [10], which havebeen shown to perform well with or without and the presence of concept drift.

In classification, we are often faced with the situation where the numbers of trainingexamples from different classes are very imbalanced. For example, in fault diagnosis orfraud detection, the majority of available examples are normal cases, the examples forfaults or frauds are minority. In such cases, applying off-the-shelf machine learningalgorithms may not lead to an appropriate learning outcome because of the bias towardthe majority class. Cost sensitive learning can be used for class imbalance learning if weknow the cost matrix of making different errors. Unfortunately, costs are very difficult todefine for many real-world problems. In this case, two alternative approaches have oftenbeen followed. One is sampling, i.e., by manipulating the training data, either over-sampling of the minority class or under-sampling of the majority class or both. The otheralternative is to design specific algorithms for class imbalance. Ensemble methods haveoften been used. However, few studies exist on the analysis of why ensembles would bea good choice and what role(s) the diversity plays. An in-depth investigation of theimpact of diversity on single class performance in multiple class imbalance learning canshed some light on the challenging problem and potential solutions [11]. One futurepossibility is to embed the strength of DyS [12], a single MLP learner for multi-classimbalance learning, into an ensemble.

Recently, online class imbalance learning has attracted more attentions fromresearchers [15]. This is more than just the combination of online learning and classimbalance learning because it is impossible to know in advance which class is amajority and which is a minority. In fact, whether a class is a minority or majoritydepends on time. A learning algorithm has to learn that. The learning algorithm alsohas to detect concept drifts and react accordingly. There are still many unsolvedproblems on this new research topic [15].

References

1. Darwen, P.J., Yao, X.: Speciation as automatic categorical modularization. IEEE Trans.Evol. Comput. 1(2), 101–108 (1997)

2. Khare, V., Yao, X., Sendhoff, B.: Multi-network evolutionary systems and automaticproblem decomposition. Int. J. Gen. Syst. 35(3), 259–274 (2006)

3. Yao, X., Liu, Y.: Making use of population information in evolutionary artificial neuralnetworks. IEEE Trans. Syst. Man Cybern. Part B: Cybern. 28(3), 417–425 (1998)

4. Liu, Y., Yao, X.: Ensemble learning via negative correlation. Neural Netw. 12(10), 1399–1404 (1999)

5. Schapire, R.E.: The strength of weak learnability. Mach. Learn. 5(2), 197–227 (1990)6. Breiman, L.: Bagging predictors. Mach. Learn. 24(2), 123–140 (1996)7. Chandra, A., Yao, X.: Ensemble learning using multi-objective evolutionary algorithms.

J. Math. Model. Algorithms 5(4), 417–445 (2006)8. Chen, H., Yao, X.: Multiobjective neural network ensembles based on regularized negative

correlation learning. IEEE Trans. Knowl. Data Eng. 22(12), 1738–1751 (2010)9. Minku, L.L., White, A., Yao, X.: The impact of diversity on on-line ensemble learning in thepresence of concept drift. IEEE Trans. Knowl. Data Eng. 22(5), 730–742 (2010)

10. Minku, L.L., Yao, X.: DDD: a new ensemble approach for dealing with concept drift. IEEETrans. Knowl. Data Eng. 24(4), 619–633 (2012)

XX X. Yao

11. Wang, S., Yao, X.: Multi-class imbalance problems: analysis and potential solutions. IEEETrans. Syst. Man Cybern. Part B 42(4), 1119–1130 (2012)

12. Lin, M., Tang, K., Yao, X.: A dynamic sampling approach to training neural networks formulti-class imbalance classification. IEEE Trans. Neural Netw. Learn.Syst. 24(4), 647–660(2013)

13. Tang, E.K., Suganthan, P.N., Yao, X.: An analysis of diversity measures. Mach. Learn. 65,247–271 (2006)

14. Brown, G., Wyatt, J.L., Harris, R., Yao, X.: Diversity creation methods: a survey andcategorisation. Inf. Fusion 6(1), 5–20 (2005)

Ensemble Approaches in Learning XXI

Contents

Invited Paper

Recent Advances in Recommender Systems and Future Directions . . . . . . . . 3Xia Ning and George Karypis

Foundations of Machine Learning

On the Number of Rules and Conditions in Mining Datawith Attribute-Concept Values and “Do Not Care” Conditions . . . . . . . . . . . 13

Patrick G. Clark and Jerzy W. Grzymala-Busse

Simplifying Contextual Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Ivo Düntsch and Günther Gediga

Towards a Robust Scale Invariant Feature Correspondence. . . . . . . . . . . . . . 33Shady Y. El-Mashad and Amin Shoukry

A Comparison of Two Approaches to Discretization:Multiple Scanning and C4.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Jerzy W. Grzymala-Busse and Teresa Mroczek

Hierarchical Agglomerative Method for Improving NPS. . . . . . . . . . . . . . . . 54Jieyan Kuang, Zbigniew W. Raś, and Albert Daniel

A New Linear Discriminant Analysis Method to Addressthe Over-Reducing Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Huan Wan, Gongde Guo, Hui Wang, and Xin Wei

Image Processing

Procedural Generation of Adjustable Terrain for Applicationin Computer Games Using 2D Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Izabella Antoniuk and Przemysław Rokita

Fixed Point Learning Based 3D Conversion of 2D Videos . . . . . . . . . . . . . . 85Nidhi Chahal and Santanu Chaudhury

Fast and Accurate Foreground Background Separationfor Video Surveillance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

Prashant Domadiya, Pratik Shah, and Suman K. Mitra

Enumeration of Shortest Isothetic Paths Inside a Digital Object. . . . . . . . . . . 105Mousumi Dutt, Arindam Biswas,and Bhargab B. Bhattacharya

Modified Exemplar-Based Image Inpainting via Primal-Dual Optimization . . . 116Veepin Kumar, Jayanta Mukhopadhyay,and Shyamal Kumar Das Mandal

A Novel Approach for Image Super Resolution Using Kernel Methods . . . . . 126Adhish Prasoon, Himanshu Chaubey, Abhinav Gupta, Rohit Garg,and Santanu Chaudhury

Generation of Random Triangular Digital CurvesUsing Combinatorial Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

Apurba Sarkar, Arindam Biswas, Mousumi Dutt,and Arnab Bhattacharya

Image Retrieval

Tackling Curse of Dimensionality for Efficient ContentBased Image Retrieval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

Minakshi Banerjee and Seikh Mazharul Islam

Face Profile View Retrieval Using Time of Flight Camera Image Analysis. . . 159Piotr Bratoszewski and Andrzej Czyżewski

Context-Based Semantic Tagging of Multimedia Data . . . . . . . . . . . . . . . . . 169Nisha Pahal, Santanu Chaudhury, and Brejesh Lall

Image Tracking

Real-Time Distributed Multi-object Tracking in a PTZ Camera Network . . . . 183Ayesha Choudhary, Shubham Sharma, Indu Sreedevi,and Santanu Chaudhury

Improved Simulation of Holography Based on Stereoscopyand Face Tracking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

Łukasz Dąbała and Przemysław Rokita

Head Pose Tracking from RGBD Sensor Based on DirectMotion Estimation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Adam Strupczewski, Błażej Czupryński, Władysław Skarbek,Marek Kowalski, and Jacek Naruniec

Pattern Recognition

A Novel Hybrid CNN-AIS Visual Pattern Recognition Engine . . . . . . . . . . . 215Vandna Bhalla, Santanu Chaudhury, and Arihant Jain

XXIV Contents

Modified Orthogonal Neighborhood Preserving Projectionfor Face Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

Purvi Koringa, Gitam Shikkenawis, Suman K. Mitra, and S.K. Parulkar

An Optimal Greedy Approximate Nearest Neighbor Method in StatisticalPattern Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

Andrey V. Savchenko

Ear Recognition Using Block-Based Principal Component Analysisand Decision Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

Alaa Tharwat, Abdelhameed Ibrahim, Aboul Ella Hassanien,and Gerald Schaefer

Data Mining Techniques for Large Scale Data

Binarizing Change for Fast Trend Similarity Based Clusteringof Time Series Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

Ibrahim K.A. Abughali and Sonajharia Minz

Big Data Processing by Volunteer Computing Supportedby Intelligent Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

Jerzy Balicki, Waldemar Korłub, and Jacek Paluszak

Two Stage SVM and kNN Text Documents Classifier . . . . . . . . . . . . . . . . . 279Marcin Kępa and Julian Szymański

Task Allocation and Scalability Evaluation for Real-Time MultimediaProcessing in a Cluster Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

Jerzy Proficz and Henryk Krawczyk

Fuzzy Computing

Concept Synthesis Using Logic of Prototypes and Counterexamples:A Graded Consequence Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303

Soma Dutta and Piotr Wasilewski

Fuzzy Rough Sets Theory Reducts for Quantitative Decisions – Approachfor Spatial Data Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314

Anna Fiedukowicz

Fuzzy Rough Sets Theory Applied to Parameters of Eye Movements CanHelp to Predict Effects of Different Treatments in Parkinson’s Patients . . . . . 325

Anna Kubis, Artur Szymański, and Andrzej W. Przybyszewski

Determining OWA Operator Weights by Maximum DeviationMinimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335

Wlodzimierz Ogryczak and Jaroslaw Hurkala

Contents XXV

Fuzzy Set Interpretation of Comparator Networks . . . . . . . . . . . . . . . . . . . . 345Łukasz Sosnowski and Dominik Ślęzak

Inverted Fuzzy Implications in Backward Reasoning . . . . . . . . . . . . . . . . . . 354Zbigniew Suraj and Agnieszka Lasek

Rough Sets

Generating Core Based on Discernibility Measure and MapReduce . . . . . . . . 367Michal Czolombitko and Jaroslaw Stepaniuk

Music Genre Recognition in the Rough Set-Based Environment . . . . . . . . . . 377Piotr Hoffmann and Bożena Kostek

Scalability of Data Decomposition Based Algorithms:Attribute Reduction Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

Piotr Hońko

Application of Fuzzy Rough Sets to Financial Time Series Forecasting . . . . . 397Mariusz Podsiadło and Henryk Rybinski

A New Post-processing Method to Detect Brain TumorUsing Rough-Fuzzy Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407

Shaswati Roy and Pradipta Maji

Rough Set Based Modeling and Visualization of the Acoustic Field Aroundthe Human Head. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418

Piotr Szczuko, Bożena Kostek, Józef Kotus, and Andrzej Czyżewski

Global Optimization of Exact Association Rules Relative to Coverage. . . . . . 428Beata Zielosko

Bioinformatics

PDP-RF: Protein Domain Boundary Prediction Using RandomForest Classifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441

Piyali Chatterjee, Subhadip Basu, Julian Zubek, Mahantapas Kundu,Mita Nasipuri, and Dariusz Plewczynski

A New Similarity Measure for Identification of Disease Genes . . . . . . . . . . . 451Pradipta Maji, Ekta Shah, and Sushmita Paul

MaER: A New Ensemble Based Multiclass Classifier for Binding ActivityPrediction of HLA Class II Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462

Giovanni Mazzocco, Shib Sankar Bhowmick, Indrajit Saha,Ujjwal Maulik, Debotosh Bhattacharjee, and Dariusz Plewczynski

XXVI Contents

Selection of a Consensus Area Size for Multithreaded Wavefront-BasedAlignment Procedure for Compressed Sequences of ProteinSecondary Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472

Dariusz Mrozek, Bożena Małysiak-Mrozek, Bartek Socha,and Stanisław Kozielski

Supervised Cluster Analysis of miRNA Expression DataUsing Rough Hypercuboid Partition Matrix . . . . . . . . . . . . . . . . . . . . . . . . 482

Sushmita Paul and Julio Vera

Analysis of AmpliSeq RNA-Sequencing Enrichment Panels . . . . . . . . . . . . . 495Marek S. Wiewiorka, Alicja Szabelska, and Michal J. Okoniewski

Consensus-Based Prediction of RNA and DNA Binding Residuesfrom Protein Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501

Jing Yan and Lukasz Kurgan

Applications of Artificial Intelligence

Fusion of Static and Dynamic Parameters at Decision Level in HumanGait Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515

Marcin Derlatka and Mariusz Bogdan

Web Search Results Clustering Using Frequent Termset Mining . . . . . . . . . . 525Marek Kozlowski

Effective Imbalanced Classification of Breast Thermogram Features . . . . . . . 535Bartosz Krawczyk and Gerald Schaefer

Rician Noise Removal Approach for Brain MR Images Using KernelPrincipal Component Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545

Ashish Phophalia and Suman K. Mitra

Climate Network Based Index Discovery for Predictionof Indian Monsoon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554

Moumita Saha and Pabitra Mitra

Using Patterns in Computer Go . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565Leszek Stanisław Śliwa

Event Detection from Business News . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575Ishan Verma, Lipika Dey, Ramakrishnan S. Srinivasan,and Lokendra Singh

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587

Contents XXVII