The study of an improved FCM clustering algorithm

There are two problems for clustering algorithm of Classic Fuzzy C-Means (FCM). First, the algorithm of FCM often obtains different clustering results with the different initial cluster centers because it is over-dependent on the initial cluster centers. Second, the algorithm needs to know the actual number of clusters in advance, but in fact the number of clusters is unknown. This paper proposes a solution that we determine a reasonable number and centers of clusters using a weighted Euclidean clustering method, and then use the classical FCM algorithm. It can be significantly reduced the number of algorithm iterations. This method was proved feasibility and effectiveness through the emulation experiment.