Title: A clustering algorithm based on an estimated distribution model

Authors: Ling Tan, David Taniar, Kate A. Smith

Addresses: School of Business Systems, Monash University, Clayton, Vic 3800, Australia. ' School of Business Systems, Monash University, Clayton, Vic 3800, Australia. ' School of Business Systems, Monash University, Clayton, Vic 3800, Australia

Abstract: This paper applies an estimated distribution model to clustering problems. The proposed clustering method makes use of an inter-intra cluster metric and performs a conditional split-merge operation. With conditional splitting and merging, the proposed clustering method does not require the information of cluster number and an improved cluster vector is subsequently guaranteed. In addition, this paper compares movement conditions between inter-intra cluster metric and intra cluster metric. It proves that, under some conditions, the intersection of convergence space between inter-intra cluster metric and intra cluster metric is not empty, and neither is the other subset in the convergence space. This sheds light on how much a cluster metric can play in clustering convergence.

Keywords: estimated distribution algorithms; clustering algorithms; clustering metric; maximum entropy; data mining; conditional splitting; conditional merging; convergence space; clustering convergence.

DOI: 10.1504/IJBIDM.2005.008364

International Journal of Business Intelligence and Data Mining, 2005 Vol.1 No.2, pp.229 - 245

Published online: 08 Dec 2005 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article