A Clustering-Ensemble Approach Based On Voting

Fanrong Meng,Xuejiao Tong,Zhixiao Wang
DOI: https://doi.org/10.1007/978-3-642-23881-9_55
2011-01-01
Abstract:Clustering-ensemble can protect private information, process distributed data and reuse of knowledge, besides, noise and outliers have little effect on clustering results. This paper proposes a new clustering ensemble algorithm, based on voting, introduce correlation to represent the similarity of clusters. The correspondence between labeled vectors can be established because clusters with lager correlation share the same cluster labels. After reunification, the labeled vectors will be used to decide the final cluster result. Analysis and experiments show that the proposed algorithm could be used to clustering-ensemble and effectively improve the clustering results.
What problem does this paper attempt to address?