RFCM clustering algorithm based on adaptive weights for radar signal sorting
Qiang Zhang, Hongwei Wang, Yuanzhi Yang, Wenzhe Wang · 2016
Rough Fuzzy C-Means (RFCM) clustering algorithm is a valid algorithm to process the inseparability of border of clusters, which handles isolated data well and reduces influence of noise on clusters results. However, different data objects are unified with fixed weights, which influences clusters results greatly. In order to solve the problem, RFCM clustering algorithm based on self-adaptive weights is proposed. According to different distance between every data object and clustering center, using improved arc cotangent function to redistribute the distance, acquiring adaptive weights through the equivalence factor in each iteration, then carrying out RFCM clustering algorithm. The simulations are performed on UCI data sets, and the results show the validity of the proposed algorithm. Furthermore, the proposed algorithm is applied in sorting radar signal, and the results show the practicability of the proposed algorithm. KEYWORD: clustering; adaptive weights; equivalence factor; radar signal sorting. 4th International Conference on Mechanical Materials and Manufacturing Engineering (MMME 2016) © 2016. The authors Published by Atlantis Press 127 dating the effectiveness of the proposed algorithm and its practical applicability on radar signal sorting. 2 RFCM CLUSTERING ALGORITHM Lingrus came up with the RCM clustering algorithm, the main idea is using the upper approximation set and the lower approximation set’s concepts of rough set to improve C-means clustering algorithm. RFCM clustering algorithm is added the weights of fuzzy membership degree based on RCM clustering algorithm, which takes the membership degree as the thin judgment criterion. At present, there are two main RFCM clustering algorithms proposed by Mitra and Maji (MITRA S, 2006, MAJI P, 2007), Maji’s RFCM clustering algorithm is adapted in this paper. Let 1 N { , , } X x x as sample data set, j x denotes a data object, i C denotes the ith classes, i C and i C denote respectively the approximation set and the lower approximation set of ith classes; 1 V { , , } k v v , i v denotes the clustering center of i C ; ij u denotes the degree of membership of jth object belong to ith classes. The calculation formula of RFCM algorithm’s clustering center is as follows: l u i i i