Homogeneousness and Density-based Cluster Validity Index

Clustering validation, despite classification, is not that straightforward. Different algorithms with different parameters can produce different clusters. Without having prior knowledge, choosing the proper algorithm and parameters would be a challenging task. Therefore, the cluster validity index(CVI) plays an essential role in cluster evaluation. Although there are many CVIs available in the literature, most of them cannot handle arbitrary shapes. In this paper, a novel cluster validity index is proposed based on the homogeneousness and density of clusters that can handle arbitrary shapes.Experiments on a total of 28 real and synthetic datasets reveal promising results, based on which the efficacy and applicability of the proposed CVI for clustering evaluation is demonstrated