Clustering aggregation by probability accumulation

Since a large number of clustering algorithms exist, aggregating different clustered partitions into a single consolidated one to obtain better results has become an important problem. In Fred and Jain's evidence accumulation algorithm, they construct a co-association matrix on original partition labels, and then apply minimum spanning tree to this matrix for the combined clustering. In this paper, we will propose a novel clustering aggregation scheme, probability accumulation. In this algorithm, the construction of correlation matrices takes the cluster sizes of original clusterings into consideration. An alternate improved algorithm with additional pre- and post-processing is also proposed. Experimental results on both synthetic and real data-sets show that the proposed algorithms perform better than evidence accumulation, as well as some other methods.