An improvement of stability based method to clustering

Abstract

In recent years, the concept of clustering stability is widely used to determining the number of clusters in a given dataset. This paper proposes an improvement of stability methods based on bootstrap technique. This amelioration is achieved by combining the instability property with an evaluation criterion and using a DCA (Difference Convex Algorithm) based clustering algorithm. DCA is an innovative approach in nonconvex programming, which has been successfully applied to many (smooth or nonsmooth) large-scale nonconvex programs in various domains. Experimental results on both synthetic and real datasets are promising and demonstrate the effectiveness of our approach.