A comparison study of clustering validity indices

Abstract

Most of the algorithms of clustering take in entry certain parameters such as the number and the density of clusters or, at least, the number of data in every cluster. The question that arises is how to determine the number of clusters in the case of an automatic classification. The purpose of our study is to compare cluster validity indices to select the optimal ones. An examination of 30 indices for determining the number of clusters is conducted on real and artificial data sets being generated according to various design factors.