Active sampling for constrained clustering

Constrained Clustering is a framework of improving clustering performance by using supervised information, which is generally a set of constraints about data pairs. Since performance of constrained clustering depends on a set of constraints to use, we need a method to select good constraints that are expected to promote clustering performance. In this paper, we propose such a method, which actively select data pairs to be constrained by using variance of clustering iteration. This method consists of a bagging based cluster ensemble algorithm that integrates a set of clusters produced by a constrained k-means with random ordered data assignment. Experimental results show that our method outperforms clustering with random sampling method.