Optimization of the number of clusters in fuzzy clustering

Clustering analysis is one of the most important research topics in unsupervised pattern recognition. Fuzzy clustering can reflect the real world more objectively since it establishes the uncertainty description from the sample to the classes. One of the main shortcomings of the traditional fuzzy clustering algorithm is that the number of clusters for reaching the optimal arrangement is not calculated automatically and needs user intervention. In this paper, by adopting the idea of hierarchical clustering, we propose a new adaptive fuzzy clustering algorithm A-FCM that can determine the optimal cluster number automatically and efficiently. Numerical experiments demonstrate that the A-FCM achieves better performance than other adaptive fuzzy clustering algorithms that determine the cluster number through various validity functions such as the Xie-Beni index.