studies in Classification, Data Analysis, and Knowledge Organization Managing Editors Editorial Board H.-H. Bock, Aachen Ph. Arable, Newark W. Gaul, Karlsruhe D. Baier, Cottbus M. Vichi, Rome E Critchley, Mihon Keynes R. Decker, Bielefeld E. Diday, Paris M. Greenacre, Barcelona C. Lauro, Naples J. Meulman, Leiden P Monari, Bologna S. Nishisato, Toronto N. Ohsumi, Tokyo O. Opitz, Augsburg G. Ritter, Passau M. Schader, Mannheim C. Weihs, Dortmund Wolfgang Gaul Daniel Baier • Reinhold Decker Lars Schmidt-Thieme Editors Data Analysis and Decision Support Foreword by Shizuhiko Nishisato ^S Prof. Dr. Daniel Baier Chair of Marketing and Innovation Management Institute of Business Administration and Economics Brandenburg University of Technology Cottbus Konrad-Wachsmann-Allee 1 03046 Cottbus Germany [email protected] Prof Dr. Reinhold Decker Chair of Marketing Department of Business Administration and Economics Bielefeld University Universitatsstr. 25 33615 Bielefeld Germany [email protected] Prof Dr. Dr. Lars Schmidt-Thieme Computer Based New Media Group (CGNM) Institute for Computer Science University of Freiburg Georges-Kohler-Allee 51 79110 Freiburg Germany [email protected] ISSN 1431-8814 ISBN 3-540-26007-2 Springer-Verlag Berlin Heidelberg New York Library of Congress Control Number: 2005926825 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way and storage in data banks. Duphcation of this pubhcation or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are hable for prosecution under the German Copyright Law Springer • Part of Springer Science+Business Media springeronhne.com © Springer-Verlag Berlin • Heidelberg 2005 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this pubhcation 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. Softcover-Design: Erich Kirchner, Heidelberg SPIN 11427827 43/3153 - 5 4 3 2 1 0 - Printed on acid-free paper Foreword It is a great privilege and pleasure to write a foreword for a book honor ing Wolfgang Gaul on the occasion of his sixtieth birthday. Wolfgang Gaul is currently Professor of Business Administration and Management Science and the Head of the Institute of Decision Theory and Management Science, Faculty of Economics, University of Karlsruhe (TH), Germany. He is, by any measure, one of the most distinguished and eminent scholars in the world today. Wolfgang Gaul has been instrumental in numerous leading research initia tives and has achieved an unprecedented level of success in facilitating com munication among researchers in diverse disciplines from around the world. A particularly remarkable and unique aspect of his work is that he has been a leading scholar in such diverse areas of research as graph theory and net work models, reliability theory, stochastic optimization, operations research, probability theory, sampling theory, cluster analysis, scaling and multivariate data analysis. His activities have been directed not only at these and other theoretical topics, but also at applications of statistical and mathematical tools to a multitude of important problems in computer science (e.g., web- mining), business research (e.g., market segmentation), management science (e.g., decision support systems) and behavioral sciences (e.g., preference mea surement and data mining). All of his endeavors have been accomplished at the highest level of professional excellence. Wolfgang Gaul's distinguished contributions are reflected through more than 150 journal papers and three well-known books, as well as 17 edited books. This considerable number of edited books reflects his special ability to organize national and international conferences, and his skill and ded ication in successfully providing research outputs with efficient vehicles of dissemination. His talents in this regard are second to none. His singular commitment is also reflected by his contributions as President of the German Classiflcation Society, and as a member of boards of directors and trustees of numerous organizations and editorial boards. For these contributions, the scientiflc community owes him a profound debt of gratitude. Wolfgang Gaul's impact on research has been felt in the lives of many re searchers in many fields in many countries. The editors of this book, Daniel Baier, Reinhold Decker and Lars Schmidt-Thieme, are distinguished former students of Wolfgang Gaul, whom I had the pleasure of knowing when they were hard-working students under his caring supervision and guidance. This book is a fitting tribute to Wolfgang Gaul's outstanding research career, for VI Foreword it is a collection of contributions by those who have been fortunate enough to know him personally and who admire him wholeheartedly as a person, teacher, mentor, and friend. A glimpse of the content of the book shows two groups of papers, data analysis and decision support. The first section starts with symbolic data analysis, and then moves to such topics as cluster analysis, asymmetric mul tidimensional scaling, unfolding analysis, multidimensional data analysis, ag gregation of ordinal judgments, neural nets, pattern analysis, Markov process, confidence intervals and ANOVA models with generalized inverses. The sec ond section covers a wide range of papers related to decision support systems, including a long-term strategy for an urban transport system, loyalty pro grams, heuristic bundling, E-commerce, QFD and conjoint analysis, equity analysis, OR methods for risk management and German business cycles. This book showcases the tip of the iceberg of Wolfgang Gaul's influence and im pact on a wide range of research. The editors' dedicated work in publishing this book is now amply rewarded. Finally, a personal note. No matter what conferences one attends, Wolf gang Gaul always seems to be there, carrying a very heavy load of papers, transparencies, and a computer. He is always involved, always available, and always ready to share his knowledge and expertise. Fortunately, he is also highly organized - an important ingredient of his remarkable success and productivity. I am honoured indeed to be his colleague and friend. Good teachers are those who can teach something important in life, and Wolfgang Gaul is certainly one of them. I hope that this book gives him some satisfaction, knowing that we all have learned a great deal from our association with him. Toronto, Canada, April 2005 Shizuhiko Nishisato Preface This year, in July, Wolfgang Gaul will celebrate his 60th birthday. He is Professor of Business Administration and Management Science and one of the Heads of the Institute of Decision Theory and Management Science at the Faculty of Economics, University of Karlsruhe (TH), Germany. He received his Ph.D. and Habilitation in mathematics from the University of Bonn in 1974 and 1980 respectively. For more than 35 years, he has been an active researcher at the interface between • mathematics, operations research, and statistics, • computer science, as well as • management science and marketing with an emphasis on data analysis and decision support related topics. His publications and research interests include work in areas such as • graph theory and network models, reliability theory, optimization, stochas tic optimization, operations research, probability theory, statistics, sam pling theory, and data analysis (from a more theoretical point of view) as well as • applications of computer science, operations research, and management science, e.g., in marketing, market research and consumer behavior, prod uct management, international marketing and management, innovation and entrepreneurship, pre-test and test market modelling, computer- assisted marketing and decision support, knowledge-based approaches for marketing, data and web mining, e-business, and recommender systems (from a more application-oriented point of view). His work has been published in numerous journals like Annals of Operations Research, Applied Stochastic Models and Data Analysis, Behaviormetrika, Decision Support Systems, International Journal of Research in Marketing, Journal of Business Research, Journal of Classification, Journal of Economet rics, Journal of Information and Optimization Sciences, Journal of Marketing Research, Marketing ZFP, Methods of Operations Research, Zeitschrift fiir Betriebswirtschaft, Zeitschrift fiir betriebswirtschaftliche Forschung as well as in numerous refereed proceedings volumes. His books on computer-assisted marketing and decision support - e.g. the well-known and wide-spread book "Computergestiitztes Marketing" (pub lished 1990 together with Martin Both) - imply early visions of the nowadays ubiquitous availability and usage of information-, model-, and knowledge- oriented decision aids for marketing managers. Equipped with a profound VIII Preface mathematical background and a high degree of commitment to his research topics, Wolfgang Gaul has strongly contributed in transforming marketing and marketing research into a data-, model-, and decision-oriented quantita tive discipline. Wolfgang Gaul was one of the presidents of the German Classification Society GfKl (Gesellschaft fiir Klassifikation) and chaired the program com mittee of numerous international conferences. He is one of the managing editors of "Studies in Classification, Data Analysis, and Knowledge Organi zation", a series which aims at bringing together interdisciplinary research from different scientific areas in which the need for handling data problems and for providing decision support has been recognized. Furthermore, he was a scientific principal of comprehensive DFG projects on marketing and data analysis. Last but not least Wolfgang Gaul has positively infiuenced the research interests and careers of many students. Three of them have decided to honor his merits with respect to data analysis and decision support by inviting colleagues and friends of him to provide a paper for this "Festschrift" and were delighted - but not surprised - about the positive reactions and the high number and quality of articles received. The present volume is organized into two parts which try to refiect the research topics of Wolfgang Gaul: a more theoretical part on "Data Anal ysis" and a more application-oriented part on "Decision Support". Within these parts contributions are listed in alphabetical order with respect to the authors' names. All authors send their congratulations ''Happy birthday, Wolfgang GauF and hope that he will be as active in his and our research fields of interest in the future as he had been in the past. Finally, the editors would like to cordially thank Dr. Alexandra Rese for her excellent work in preparing this volume, all authors for their cooperation during the editing process, as well as Dr. Martina Bihn and Christiane Beisel from Springer-Verlag for their help concerning all aspects of publication. Cottbus, Bielefeld, Freiburg Daniel Baier April 2005 Reinhold Decker Lars Schmidt-Thieme Contents Part I. Data Analysis Optimization in Symbolic Data Analysis: Dissimilarities, Class Centers, and Clustering 3 Hans-Hermann Bock An Efficient Branch and Bound Procedure for Restricted Principal Components Analysis 11 Wayne S, DeSarho, Robert E. Hausman A Tree Structured Classifier for Symbolic Class Description .. 21 Edwin Diday, M. Mehdi Limam, Suzanne Winsberg A Diversity Measure for Tree-Based Classifier Ensembles 30 Eugeniusz Gatnar Repeated Confidence Intervals in Self—Organizing Studies .... 39 Joachim Hartung, Guido Knapp Fuzzy and Crisp Mahalanobis Fixed Point Clusters 47 Christian Hennig Interpretation Aids for Multilayer Perceptron Neural Nets ... 57 Harald Hruschka An Unfolding Scaling Model for Aggregated Preferential Choice Data 65 Tadashi Imaizumi Model-Based Clustering - Discussion on Some Approaches ... 73 Krzysztof Jajuga Three—Way Multidimensional Scaling: Formal Properties and Relationships Between Scaling Methods 82 Sabine Krolak-Schwerdt Empirical Approach as a Scientific Framework for Data Analysis 91 Shizuhiko Nishisato Asymmetric Multidimensional Scaling of Relationships Among Managers of a Firm 100 Akinori Okada, Tadashi Imaizumi, Hiroshi Inoue X Contents Aggregation of Ordinal Judgements Based on Condorcet's Majority Rule 108 Otto Opitz, Henning Paul ANOVA Models with Generalized Inverses 113 Wolfgang Polasek, Shuangzhe Liu Patterns in Search Queries 122 Nadine Schmidt-Mdnz, Martina Koch Performance Drivers for Depth-First Frequent Pattern Mining 130 Lars Schmidt-Thieme, Martin Schader On the Performance of Algorithms for Two-Mode Hierarchical Cluster Analysis — Results from a Monte Carlo Simulation Study 141 Manfred Schwaiger, Raimund Rix Clustering Including Dimensionality Reduction 149 Maurizio Vichi The Number of Clusters in Market Segmentation 157 Ralf Wagner, Soren W. Scholz, Reinhold Decker On Variability of Optimal Policies in Markov Decision Processes 177 Karl-Heinz Waldmann Part II. Decision Support Linking Quality Function Deployment and Conjoint Analysis for New Product Design 189 Daniel Baier, Michael Brusch Financial Management in an International Company: An OR- Based Approach for a Logistics Service Provider 199 Ingo Bockenholt, Herbert Geys Development of a Long-Term Strategy for the Moscow Urban Transport System 204 Martin Both The Importance of E-Commerce in China and Russia — An Empirical Comparison 212 Reinhold Decker, Antonia Hermelbracht, Frank Kroll Analyzing Trading Behavior in Transaction Data of Electronic Election Markets 222 Markus Franke, Andreas Geyer-Schulz, Bettina Hoser