On post-clustering evaluation and modification

Unsupervised clustering algorithms sometimes do not lead to meaningful inte rpretations of the structure in the data. We propose a new approach in whic h the concept of cluster density is introduced to assess the quality of an algorithmically generated partition and accordingly guide an amelioration p rocess through split-and-merge operations. (C) 2000 Published by Elsevier S cience B.V. All rights reserved.