A Density Discriminant Index for Cluster Validation

Abstract

Clustering analysis is widely applied in several domains of study. Using a suitable number of clusters is one of the most important factors to influence the performance of clustering. Several algorithms of cluster validation have been developed to find such a number. In this paper, we proposed a method for cluster validation adapted from the Discrimination Evaluation via Optic Diffraction Analysis (DEODA) algorithm to derive an appropriate number of clusters. In particular, our method uses DEODA to perform within- and between-cluster discrimination analysis in order to find the suitable number of clusters. We evaluate our method by comparing similarity score against the existing cluster validation algorithm i.e., the Silhouette index. The results show that the similarity scores derived from our method are higher than results yielded from the Silhouette index.