Wolfgang Gaul

Between Data Science and Applied Data Analysis

The volume presents new developments in data analysis and classification and gives an overview of the state of the art in these scientific fields and relevant applications. Areas that receive considerable attention in the book are clustering, discriminitation, data analysis, and statistics, as well as applications in economics, biology, and medicine. The reader will find material on recent technical and methodological developments and a large number of application papers demonstrating the usefulness of the newly developed techniques.

Mining Web Navigation Path Fragments

For many web usage mining applications like, e.g., user segmentation, it is crucial to compare navigation paths of different users. We model user navigation path fragments by generalized subsequences that take into consideration local deviations but still sketch the global user navigational behavior. This paper presents a new algorithm of apriori type for mining all generalized subsequences of user navigation paths with prescribed minimal occurrence from a given database.

A New Algorithm for Two-Mode Clustering

Against the background that one-mode clustering, which is based on similarity or dissimilarity data, is well known and widely used, quite a number of generalizations of the basic clustering methodology have been developed. For so-called two-mode data we report on research within the area of two-mode clustering and describe the new AE (Alternating Exchanges) algorithm. For non- overlapping two-mode clustering this algorithm is based on the exchange of the cluster membership of elements from the sets of different modes. Therefore, it is simple and very fast. The results of applying this algorithm to several concrete data sets are compared to those of an own PENCLUS (PENalty CLUStering) approach.

Two-Mode Overlapping Clustering With Applications to Simultaneous Benefit Segmentation and Market Structuring

A new two-mode overlapping clustering procedure is presented. This procedure includes solution possibilities for two-mode (non-)overlapping additive clustering as well as (non-)overlapping clusterwise regression with conjoint experiments and can be used for simultaneous benefit segmentation and market structuring. Applications of various cases of the new procedure to conjoint data are used for comparisons.

Optimal product positioning based on paired comparison data

A new approach for analyzing paired comparison data is proposed which combines a probabilistic ideal point model with product positioning issues. Unlike traditional approaches based on paired comparison data the same formulation is used for estimating a joint space representation of consumer segments and products as well as for determining optimal (new) product positioning options in a relevant product-market. A Monte Carlo experiment is presented and real-world coffee market data are used to show advantages of the new approach.

Web Robot Detection - Preprocessing Web Logfiles for Robot Detection

Web usage mining has to face the problem that parts of the underlying logfiles are created by robots. While cooperative robots identify themselves and obey to the instructions of server owners not to access parts or all of the pages on the server, malignant robots may camouflage themselves and have to be detected by web robot scanning devices. We describe the methodology of robot detection and show that highly accurate tools can be applied to decide whether session data was generated by a robot or a human user.

Classification and Information Processing at the Turn of the Millennium

Data Analysis and Classification: G. Arminger, J. Wittenberg, P. Stein: New Developments in Latent Variable Models.- H.-H. Bock: Regression-Type Models for Kohonens Self-Organizing Networks.-J. Meulman: Relational Data Analysis, Discrimination and Classification.- S. Nishisato: Data Types and Information: Beyond the Current Practice of Data Analysis.-A. Okada, T. Imaizumi: Two-Mode Three-Way Asymmetric Multidimensional Scaling with Constraints on Asymmetry.-H. Bensmail, J. Meulman: Regularized Discriminant Analysis with Optimally Scaled Data.-W. Castillo, J. Trejos: Recurrence Properties in Two-Mode Hierarchical Clustering.-J. Conde, M. Colomer, C. Capdevila, A. Gil: On Pyramidal Classification in the Delimitation of Climatic Regions in Catalonia.-F. Dau, R. Wille: On the Modal Understanding of Triadic Contexts.-S. Gabler, J. Blasius: Clustering and Scaling: Grouping Variables in Burt Matrices.-K. Jajuga, M. Walesiak: Standardisation of Data Set under Different Mesurement Scales.-P. Lory, D. Gietl: Neural Networks for Two-Group Classification Problems with Monotonicity Hints.-R. Pfluger, O. Gefeller: A Bridge from the Nearest Neighbour to the Fixed Bandwidth in Nonparametric Functional Estimation.-W. Theis, C. Weihs: Clustering Techniques for the Detection of Business Cycles.- J. Trejos, W. Castillo: Simulated Annealing Optimization for Two-mode Partitioning.-Computer Science, Computational Statistics, and Data Mining: W. Hauke: Competiveness of Evolutionary Algorithms.-J. Kacprzyk: Intelligent Data Analysis via Linguistic Data Summaries: A fuzzy Logic Approach.-G. Lindner, R. Studer: Algorithm Selection Support for Classification.-K. Mainzer: Society of Knowledge - Interdisciplinary Perspectives of Information Processing in Virtual Networks.-P. Naeve: What is Computational Statistics?.-M.Feldmann: A Development Framework for Nature Analogic Heuristics.-C. Heidsiek, W. Uhr: Systematizing and Evaluating Data Mining Methods.-T. Hermes, A. Miene, P. Kreyenhop: On Textures: A Sketch of a Texture-Based Image Segmentation Approach.-T. Hermes, A. Miene, O. Moehrke: Automatic Texture Classification by Visual Properties.-S. Kuhlins, A. Korthaus: Java Servlets versus CGI - Implications for Remote Data Analysis.-D. Pechtel, K.-D. Kuhnert: Generating Automatically Local Feature Groups of Complex and Deformed Objects.-F. Sauberlich, W. Gaul: Decision Tree Construction by Association Rules.-B. Stein, O. Niggemann, U. Husemeyer: Learning Complex Similarity Measures.-M. Thomas, R.-D. Reiss: Graphical Programming in Statistics: The XGPL Prototype.-Management Science, Marketing, and Finance: U. Wagner, C. Boyer: Measuring Brand Loyalty on the Individual Level: a Comparative Study.-U. Bankhofer: Facility Location Planning with Qualitative Location Factors.-K. Bartels, Y. Boztug, M. Muller: Testing the Multinomial Logit Model.-T. Burkhardt: Time Optimal Portfolio Selection: Mean-variance-efficient Sets for Arithmetic and Geometric Brownian Price Processes.-N. Gorz, L. Hildebrandt, D. Annacker: Analyzing Multigroup Data with Structural Equation Models.-S. Hensel-Borner, H. Sattler: Validity of Customized and Adaptive Hybrid Conjoint Analysis.-Y. Kimura, A. Okada: Analysis of the Assessments of Ones Values Among Different Cohorts in Age and Sex by Multidimensional Scaling.-J. Limperger: Impacts of Hedging with Futures on Optimal Production Levels.-B. Nietert: How to Integrate Stock Price Jumps into Portfolio Selection.-W. Polasek, L. Ren: A Multivariate GARCH-M Model for Exchange Rates in the US, Germany and Japan.-A. Tangian: A Model for Constructing Quadratic

A knowledge-based system for supporting data analysis problems☆

Abstract Scientists from different research areas have developed a variety of models and methods to support data analysis problems in their specific fields of interest. However, the preconditions for a proper usage of the corresponding software which is often provided in the shape of software packages or individual programs may cause a severe problem. These preconditions preponderantly demand knowledge as well about theoretical and software specific aspects of algorithms used as about essentials of the area of application. The prototype to be presented in this paper aims at extending the capabilities of conventional software approaches by incorporating knowledge concerning the suitability of certain data analysis methods. The knowledge-based part is realized in PROLOG. The application area is market research. The prototype already comprises several data analysis procedures, especially from multidimensional scaling and cluster analysis, and data management facilities which can be invoked by the user based on the recommendations given by the system.

Mining generalized association rules for sequential and path data

While association rules for set data use and describe relations between parts of set valued objects completely, association rules for sequential data are restricted by specific interpretations of the subsequence relation: contiguous subsequences describe local features of a sequence valued object, noncontiguous subsequences its global features. We model both types of features with generalized subsequences that describe local deviations by wild cards, and present a new algorithm of a priori type for mining all generalized subsequences with prescribed minimum support from a given database of sequences. Furthermore we show that the given algorithm automatically takes into account an eventually underlying graph structure, i.e., is applicable to path data also.

Market Simulation Using a Probabilistic Ideal Vector Model for Conjoint Data

In commercial applications of conjoint analysis to product design and product pricing it has become quite popular to further evaluate the estimated individual part-worth functions by predicting shares of choices for alternatives in hypothetical market scenarios (Wittink, Vriens and Burhenne 1994 and Baier 1999 for surveys on commercial applications). Wide-spread software packages for conjoint analysis (Sawtooth Software’s 1994 ACA system) already include specific modules to handle this so-called market simulation situation for which, typically, a threefold input is required: (I) The (estimated) individual part-worth functions have to be provided. (II) A definition of a hypothetical market scenario is needed that allows to calculate individual utility values for each available alternative. (III) A so-called choice rule has to be selected, which relates individual utility values to expected individual choice probabilities and, consequently, to market shares for the alternatives. In this context, the determination of an adequate choice rule seems to be the most cumbersome task. Well-known traditional choice rules are, e.g., the 1ST CHOICE rule (where the individuals are assumed to always select the choice alternative with the highest utility value), the BTL (Bradley,Terry, Luce) rule (where individual choice probabilities are related to corresponding shares of utility values), and the LOGIT rule (where exponentiated utility values are used). Furthermore, in newer choice rules implemented by various software developers, the similarity of an alternative to other alternatives is taken into account as a corrective when choice probabilities are calculated (Sawtooth Software 1994).