ABSTRACT

In this paper, we extensively study about the important aspect of various Clustering techniques, the cluster quality. The goodness of clustering is measured in terms of cluster validity indices where the results of clustering are validated every time to give the maximum efficiency. The quality of clusters is measured in a decision-theoretic rough set oriented approach rather than the traditional geometry-based measures. Experiments are carried out with synthetic, standard and real world data for evaluating rough and crisp clustering. Also a new advancement in estimating the number of clusters in the analysis of gene expression data is studied. Here we follow a scheme called System Evolution to estimate the number of clusters based on Partitioning around medoids algorithm.