Optimum coordinate number of clusters and best clustering in fuzzy C-means

A coordinate function of criteria on the basis of intra- and inter-distances in the fuzzy C-means (FCM) is proposed. Iterative self-organizing data analysis technique algorithm (ISODATA) and discrete particle swarm optimization (PSO) are combined to form a PSO self-organizing data analysis technique algorithm (PSO-ISODATA), which is used to conduct the optimal computing of FCM. Compared to other methods, our method can be used not only to do optimal clustering but also to yield the optimum coordinate number of clusters and the corresponding optimal clustering without artificial interference according to the clustering criteria, given a preset number of clustering. PSO-ISODATA has a wide application. When other cluster criteria are adopted, only the fitness function is needed to be modified.