A Clustering Algorithm Based on Cluster Feature Weighting Fuzzy Compactness and Separation

For the clustering of imbalanced data, we propose a clustering algorithm based on cluster feature weighting fuzzy compactness and separation (CFWFCS). By addressing the deficiency of the fuzzy membership formula in the fuzzy compactness and separation (FCS) algorithm, we provide formulations of sample membership and attribute weighting for every cluster of CFWFCS, and then discuss their adjustments. The proposed CFWFCS-based method is compared with the FCS algorithm and two other weighted-clustering algorithms on benchmark datasets. Experimental results show that the proposed algorithm outperforms the other three algorithms in accuracy and reasonability for unbalanced-distribution data.