Cluster ensemble construction with the algorithm of averaged centroids

The task of finding consensus solution of cluster analysis problem is considered in the paper. A heuristic algorithm for constructing consensus clustering partition using any centroid-based algorithm is proposed. It is theoretically proved that the algorithm is statistically stable. The novelty of the algorithm is that it is implemented and optimized for running in parallel and distributed environment. The paper includes the results of testing the algorithm on artificial and real data.