A Novel FCM´s Initial Parameters Acquisition Method

The fuzzy c-means method (FCM) is an important research topic that has practical applications in many fields. However, the greatest disadvantage of this method is its sensitivity to initialized values including the number of clusters and cluster centroids. The main goal of this paper is to present an effective method for automatically determining the number of clusters and acquiring the corresponding initial centroids at the same time. This paper first uses the data reduction technique to acquire a simple cell distribution of the data set and then new techniques named cell diffusion and cell abrasion based on the cell dentist values are used to acquire a stable and high density data mass which eases the cluster detection process in the last stage. Then, the number of the centers and their corresponding centroids are acquired for FCM initialization. Experimental results and comparisons are given to illustrate the performance of the new method