A Survey on K-means Based Consensus Clustering
Nimesha M. Patil, Dipak V. Patil · 2016
Cluster ensemble techniques aim at combining multiple individual clustering solutions into a consensus one, which agrees as much as possible with existing individual clustering solutions. These individual clustering solutions may be heterogeneous in nature as they are obtained by multiple runs of different clustering algorithms or multiple runs of same algorithm with dynamic variable settings on the same dataset. There are two import key issues in designing methodology for consensus clustering problem to work on heterogeneous partitions. One is availability of good consensus function that fixes the consensus partition by verifying utilities of available existing clustered partitions. Another is use of the efficient suitable clustering methodology to fit into consensus clustering like k-means algorithm. K-means based consensus clustering (KCC) is one of the prominent solution studied in recent years of research. Hence, our survey here tries to cover recent and major advances in k-means and consensus clustering separately. Different approaches used by researchers to improve the results of both paradigms individually are promising. Picking up innovative solutions from both k-means and consensus clustering can build better integrated and sophisticated KCC frameworks. This survey can give better future directions for improving clustering quality in heterogeneous environments through the means of KCC.