Relative entropy collaborative fuzzy clustering method

• In many clustering methods, it is assumed that the data are stored in one data site.
• Ways of communication between data sites effects the clustering results.
• A new communication method based on relative entropy is proposed.
• It minimizes the divergence of membership degrees of Kth data in ith cluster in all data sites.
• RECFC and its two modes could efficiently classify data stored in several data sites.

The main task of clustering methods, especially fuzzy methods, is to find whether natural grouping exists in data and to impose identity on them. In some situations, data are stored in several data sites and to discover the global structures, clustering methods have to be aware of dependencies in all data sites. Collaborative fuzzy clustering methods have been proposed and widely studied to answer such need. In this paper, a novel collaborative fuzzy clustering method is proposed. In this method, relative entropy concept is used as the communication method, a new approach is applied to calculate the interaction coefficient between data sites, and horizontal and vertical modes of the proposed method are discussed. Performance of the proposed method is evaluated using several experiments and the results show that it has the highest quality of collaboration and could classify data more efficiently.