On Instance Weighted Clustering Ensembles

Abstract

Ensemble clustering is a technique which combines multipleclustering results, and instance weighting is a technique which highlightsimportant instances in a dataset. Both techniques are known to enhanceclustering performance and robustness. In this research, ensembles andinstance weighting are integrated with the spectral clustering algorithm.We believe this is the first attempt at creating diversity in the generativemechanism using density based instance weighting for a spectral ensemble.The proposed approach is empirically validated using synthetic datasetscomparing against spectral and a spectral ensemble with random instanceweighting. Results show that using the instance weighted sub-samplingapproach as the generative mechanism for an ensemble of spectral cluster-ing leads to improved clustering performance on datasets with imbalancedclusters.