Measuring Hybrid SC-FCM Clustering with Cluster Validity Index

Abstract

Clustering classifies data into groups based on the similarity of each element of data. In order to validate the cluster, cluster validity index is introduced. Hybrid SC-FCM (Subtractive Clustering-Fuzzy C-Means) clustering method is a clustering technique to overcome the weakness of the FCM (Fuzzy C-Means) clustering. While the hybrid SC-FCM is a promising method, no validity measurement on the resulted cluster has been done. This research measures the cluster validity index of Hybrid SC-FCM method. The cluster validity indices used in the research are partition coefficient, partition entropy, and Xen Beni Index. The research shows mix results. Even though the Hybrid SC-FCM method fails to find the best number of clusters as suggested, it shows that hybrid SC-FCM able to exceed the traditional FCM method in providing initial centroids.