Soft Cluster Ensemble Based on Fuzzy Similalrity Measure

Cluster ensemble has emerged as a powerful method for overcoming instabilities in unsupervised clustering solutions. Recent research mostly focused on the combination of crisp clusterings based on co-association matrix. Co-association matrix is generated to summarize the ensemble, and then a consensus function is devised to get the final result. In this paper, we propose a method to combine soft clusterings. Firstly, Fuzzy co-association matrix based on fuzzy similarity measure is generated to summarize the ensemble of soft clusterings. Three different fuzzy similarity measures are mentioned here. Then, multiple soft clusterings are combined by selected consensus function. Finally, experiments are performed to assess the proposed method and it shows promising results compared to general cluster ensemble methods based on crisp clusterings.