Parallelizing the unsupervised k-windows clustering algorithm

Abstract

Clustering can be defined as the process of partitioning a set of patterns into disjoint and homogeneous meaningful groups, called clusters. The growing need for parallel clustering algorithms is attributed to the huge size of databases that is common nowadays. This paper presents a parallel version of a recently proposed algorithm that has the ability to scale very well in parallel environments mainly regarding space requirements but also gaining a time speedup. © Springer-Verlag Berlin Heidelberg 2004.