Ensemble Clustering via Adaptive Weight Adjustment and Cluster Generation in Difference Similarity
Xiaoning Wang, Chun Liu, Hengshan Zhang, Yun Wang · 2023
In a clustering ensemble, determining the weight of basin clustering is a critical issue. The weight of basin clustering affect the integration of similarity matrices, which influences the final clustering result. Existing methods often suffer from poor weight assignment effectiveness, high sparsity and low value density of the similarity matrix, and difficulties in generating clustering result, which can limit the performance of the method. We propose a cluster ensemble method via adaptive weight adjustment and difference similarity for addressing these issues. First, we generate basin clustering through multiple clustering methods and adjust the weight of each basin clustering using an adaptive weight adjustment algorithm to integrate similarity matrix. Then, we use the shortest distance algorithm to reduce sparsity of the similarity matrix. In particular, the difference similarity between data point and cluster is defined for clustering result generation. Experimental results on multiple public data sets show that the proposed method outperforms existing representative clustering methods.