An Exploratory Study of the Inputs for Ensemble Clustering Technique as a Subset Selection Problem

Abstract

Ensemble and Consensus Clustering address the problem of unifying multiple clustering results into a single output to best reflect the agreement of input methods. They can be used to obtain more stable and robust clustering results in comparison with a single clustering approach. In this study, we propose a novel subset selection method that looks at controlling the number of clustering inputs and datasets in an efficient way. The authors propose a number of manual selection and heuristic search techniques to perform the selection. Our investi‐ gation and experiments demonstrate very promising results. Using these techni‐ ques can ensure better selection methods and datasets for Ensemble and Consensus Clustering and thus more efficient clustering results.