Transfer Evidential C-Means Clustering

Abstract

Clustering is widely used in text analysis, natural language processing, image segmentation and other data mining fields. ECM (evidential c-means) is a powerful clustering algorithm developed in the theoretical framework of belief functions. Based on the concept of credal partition, it extends those of hard, fuzzy, and possibilistic clustering algorithms. However, as a clustering algorithm, it can only work well when the data is sufficient and the quality of the data is good. If the data is insufficient and the distribution is complex, or the data is sufficient but polluted, the clustering result will be poor. In order to solve this problem, using the strategy of transfer learning, this paper proposes a transfer evidential c-means (TECM) algorithm. TECM employs the historical clustering centers in source domain as the reference to guide the clustering in target domain. In addition, the proposed transfer clustering algorithm can adapt to situations where the number of clusters in source domain and target domain is different. The proposed algorithm has been validated on synthetic and real-world datasets. Experimental results demonstrate the effectiveness of transfer learning in comparison with ECM and the advantage of credal partition in comparison with TFCM.