A Classification of Cluster Validity Indexes Based on Membership Degree and Applications

摘要

With the widely used of data mining and cluster analysis, cluster validation is attracting increasing attention. In this paper, the concept and development of cluster validation are introduced, then, based on the membership degree, a classification of cluster validity indexes is proposed: cluster validity indexes fit for crisp cluster, cluster validity indexes fit for fuzzy cluster. Based on this, combining with Cluster Validity Analysis Platform (CVAP), describing the two most important usages of cluster validation: to find the optimal number of clusters and to find appropriate clustering algorithms to a particular data set. Experiments give visualization representation of clustering validation process.