Concept Based Neighbor Cluster Ensemble Re-Clustering Method

Abstract

Clustering Ensemble combines the several partitions generated by different clustering algorithm into single clustering solution. The optimization-based method is proposed for the combination of cluster ensembles for the class of problems with intracluster criteria, such as Minimum-Sum-of-Squares-Clustering (MSSC). To find the solution for MSSC problem we are using simple and efficient algorithm called improved Exact Method for cluster ensemble re-clustering algorithm which uses similarity measures and distance between the weak clusters. The solution obtained by the single clustering algorithm does not provide better solution. The solution obtained by this algorithm guarantees better solutions than the ones in the individual cluster. For the MSSC problem in particular, a prototype implementation of improved Exact Method for cluster ensemble algorithm will produce a new better solution. The algorithm is particularly effective when the number of clusters is large, in which case it is able to escape the local minima found by K-means type algorithms by recombining the solutions in a Set-Covering context. The stability of the algorithm is also establish by running this algorithm several times for the same clustering problem instance, produce high-quality solutions. Finally, in experiments utilize external criteria to compute the validity of clustering. The algorithm is capable of producing high-quality results that are comparable in quality to those of the best known clustering algorithms.