Improving the Orthogonal Range Search kwindows clustering algorithm.pdf

Improving the Orthogonal Range Search -windows Algorithm P. Alevizos B. Boutsinas Department of Mathematics, Department of Business Administration, University of Patras (UOP) UOP, UPAIRC University of Patras Artificial GR-26500 Patras, Greece Intelligence Research Center (UPAIRC) vutsinas@bma.upatras.gr GR-26500 Patras, Greece alevizos@math.upatras.gr D. Tasoulis M.N. Vrahatis Department of Mathematics, UOP, UPAIRC Department of Mathematics, UOP, UPAIRC GR-26500 Patras, Greece GR-26500 Patras, Greece dtas@math.upatras.gr vrahatis@math.upatras.gr Abstract sentatives of the clusters, the conjunction of attribute values describing each mean can be considered as a clustering rule Clustering, that is the partitioning of a set of patterns for describing data (taking, of course, into consideration and into disjoint and homogeneous meaningful groups (clus- a number of certain properties as density, variance, shape ters), is a fundamental process in the practice of science. and separation [2]). Recently, the task of extracting knowl- -windows is an efficient clustering algorithm that reduces edge from large databases, in the form of clustering rules, the number of patterns that need to be examined for similar- has become