An Exponential Cluster Validity Index for Fuzzy Clustering with Crisp and Fuzzy Data
M.H. Fazel Zarandi, Mohammad Reza Faraji, Mahdi Karbasian · Scientia Iranica · 2010
This paper presents a new cluster validity index for finding a suitable number of fuzzy clusters with crisp and fuzzy data. The new index, called the ECAS-index, contains exponential compactness and separation measures. These measures indicate homogeneity within clusters and heterogeneity between clusters, respectively. Moreover, a fuzzy c-mean algorithm is used for fuzzy clustering with crisp data, and a fuzzy k-numbers clustering is used for clustering with fuzzy data. In comparison to other indices, it is evident that the proposed index is more e ective and robust under different conditions of data sets, such as noisy environments and large data sets.