A Method to Improve the Analysis of Cluster Ensembles

Abstract

Clustering is fundamental to understand the structure of data. In the past decade the cluster ensembleproblem has been introduced, which combines a set of partitions (an ensemble) of the data to obtain a singleconsensus solution that outperforms all the ensemble members. However, there is disagreement about which arethe best ensemble characteristics to obtain a good performance: some authors have suggested that highly differentpartitions within the ensemble are beneï¬ cial for the ï¬ nal performance, whereas others have stated that mediumdiversity among them is better. While there are several measures to quantify the diversity, a better method toanalyze the best ensemble characteristics is necessary. This paper introduces a new ensemble generation strategyand a method to make slight changes in its structure. Experimental results on six datasets suggest that this isan important step towards a more systematic approach to analyze the impact of the ensemble characteristics onthe overall consensus performance.