A novel automatic fuzzy clustering algorithm based on soft partition and membership information

In the field of data clustering, finding the cluster numbers automatically and generating reliable clusters for a given dataset are fundamental but challenging tasks. Recently, a clustering analysis algorithm for the automatic identification of cluster numbers was presented, and it achieves accurate clustering results for those datasets with complex structures. Unfortunately, this algorithm utilizes a hard partition approach in the process of integration and does not make full use of the membership information from each fuzzy c-means (FCM) clustering result. Thus, this scheme has to integrate many more FCM clustering results and also requires many iterations during the process of iterative graph partitioning. To address this problem, an automatic fuzzy clustering algorithm is proposed in this paper, combining the soft partition method with the membership information from each FCM clustering result. Finally, extensive experiments are performed, and under the premise of obtaining accurate clustering results simultaneously, the proposed algorithm can effectively decrease the number of FCM clustering results in the process of integration compared with the original algorithm. Furthermore, the number of iterations of the proposed scheme in the iterative graph partitioning process is half that of the original approach.