A Method of Clustering Ensemble Based on Grey Relation Analysis

Abstract

Clustering ensemble algorithm is an effective way to improve the accuracy, stability and robustness of clustering results. It can get better results by fusing multiple homogenous or heterogeneous base clustering models. In this paper a clustering ensemble algorithm based on grey relation analysis is proposed. Through constructing a grey-linked matrix, the relationship between the data objects and all clusters can be connected, then the basic clustering results can be integrated. After that the appropriate consensus function is used to get the integrated clustering results by partitioning the matrix finally. In contrast to other clustering ensemble models, the proposed algorithm has higher accuracy and gets better stability and robustness after validating on several datasets from UCI. Moreover, the proposed algorithm can effectively avoid the category labels’ matching problem that’s in traditional clustering ensemble algorithms, so the clustering performance has been greatly improved.