DEVELOPMENT OF A WEIGHTED FUZZY C-MEANS CLUSTERING ALGORITHM BASED ON JADE

Abstract. To overcome the shortcomings of falling into local optimal solutions and being too sensitive to initial values of the traditional fuzzy C-mean clustering algorithm, a weighted fuzzy C-means (FCM) clustering algorithm based on adaptive differential evolution (JADE) is proposed in this paper. To consider the particular contributions of different features, a ReliefF algorithm is used to assign the weight for each feature. A weighted morphology-similarity distance (WMSD) based on ReliefF instead of the Euclidean distance is used to improve the objective function of the FCM clustering algorithm. Experimental results on the international standard Iris data and the contrast experimental results with other evolution algorithms show that the proposed algorithm has higher clustering accuracy and greater searching capability. Key words. Fuzzy C-means, adaptive differential evolution, weighted morphology-similarity distance, clustering precisio