Selective Affinity Propagation Ensemble Clustering

Abstract

An ensemble clustering algorithm based on affinity propagation clustering (APC) algorithm is proposed in the paper. Different base clustering results are obtained by using APC algorithm under the selected preference parameter. However, some base clusters have little effect on the final clustering result, the "good" base clusters should be selected from all base clusters. Then the diversity and coverage ratio are applied to the judgement of the homogeneity of generated base clusters. Finally, the evidence accumulation clustering (EAC) algorithm is used for obtaining the co-association matrix from the selected base clusters, and generating final the clustering result. Experimental results on two data sets demonstrate that the clustering ensemble based on the refined co-association matrix outperforms some state-of-the-art clustering ensemble schemes.