Voting affinity propagation algorithm for clustering XML documents
Xu Wang, Jinmao Wei, Bao-Quan Fan, Ting Yang · 2012
Affinity propagation algorithm(AP) is a new powerful tool for unsupervised clustering. Yet AP can't work well as expected on the datasets with high-similarity among some points. All these data points are likely to become exemplars so as to generate overmuch clusters. In this paper, we introduce a new algorithm called voting affinity propagation extended from original affinity propagation(OAP) to overcome the problem of generating overmuch clusters when AP is applied to XML document clustering. We propose to select high quality exemplars from candidate exemplars based on a voting mechanism. We adopt both real and synthetic datasets to test the performance of the proposed algorithm. Experiment results show that voting affinity propagation leads to more stable clustering and works better than OAP in clustering data with extremely high similarities.