A HYBRID FUZZY C-MEDOIDS CLUSTERING USING THE WHALE OPTIMIZATION ALGORITHM

Abstract

Although the fuzzy c-means algorithm (FCM) has been widely used in many fields, they are sensitive to noise and outliers. Recently, the Fuzzy C-Medoids (FCMdd) algorithm has been shown to be more effective in dealing with noise data. The difference between FCM and FCMdd is the formation mechanism of clusters. At the same time, FCM builds clusters based on membership function and samples in the cluster. FCMdd selects some of the existing actual samples as cluster medoids. This results in FCMdd being able to handle noise better than FCM. This paper presents a hybrid approach of the whale optimization algorithm WOA with FCMdd (FWCMdd) to optimize the clustering process. This hybridization prevents FWCMdd from falling into the local trap and rapidly converging. This solution has been compared with the Fuzzy k-Medoids (FKM) algorithm and the primitive FCMdd. The results indicate that the proposed method is better than most of the evaluation indicators.