FlexiMine- A Flexible Platform for KDD Research and Application Development.pdfVIP

FlexiMine - A Flexible Platformfor KDD Researchand Application DevelopmentR. Ben-Eliyahu-Zohary, C. Domshlak, E. GudesN. Liusternik, A. Meisels, T. Rosen, S. E. ShimonyDept. of Mathematics and Computer Science,Ben-Gurion UniversityP.O. Box 653, 84105 Beer-Sheva, ISRAELAbstractFlexiMine is a KDD system designed as a testbed for data-mining research, aswell as a generic knowledge discovery tool for varied database domains. Flexibilityis achieved by an open-ended design that enables integration of existing data-miningalgorithms, new locally developed algorithms, and utility functions such as visualizationand preprocessing. The system interfaces an INFORMIX database server via SQLqueries and thus its support for new databases is simple and clean.With a view of serving both remote and local users, Internet availability was adesign goal. Fleximine is implemented in Java and hence we can run the user-end ofthe system either as a Java application or (with some limitations on the user) as a Javaapplet.This paper reviews the architecture, design, and operation of FlexiMine and presentsthe new ideas incorporated in Fleximines data-mining algorithms (Association rules,Decision trees, Bayesian knowledge-bases and Meta-queries).Keywords: Knowledge Discovery in Databases, Data-Mining, Meta-queries, Flexible Soft-ware Systems, WWW-Based Tools, Association Rules, Bayesian Knowledge Bases.This is an expanded and revised version of a short paper that was published in the KDD-98 proceedings 1 IntroductionWith the enormous growth of accessible information around us, Data Mining and KnowledgeDiscovery (KDD) have become a vital technology [10]. FlexiMine is a prototype KDD system,currently being developed at Ben-Gurion University, for testing and evaluating techniquesand algorithms for Data Mining and Knowledge Discovery. The emphasis of this system andits software architecture is on the integration of many KDD operations (including databaseaccess and selection, preprocessing, data tran