WCSE 2017
ISBN: 978-981-11-3671-9 DOI: 10.18178/wcse.2017.06.019

Kernel FCM Based on Simulated Annealing Algorithm

Chengwei Jiang, Xiaoyan Cui, Kaijian Zhang, Tongliang Li

Abstract— As everyone knows,birds of a feather flock together,which reveals the essence of clustering.Fuzzy c-means clustering algorithm(FCM) in many algorithms of clustering play a significant role in unsupervised learning.However,FCM still has the following three defects—unknown number of clusters,unstable process of clustering and discoveries of outliers . Aimed at these defects existed in the FCM,a kernel based fuzzy c-means clustering algorithm is proposed to optimize conventional FCM,based on simulated annealing algorithm which takes advantage of simulated annealing algorithm(SA) and kernel technology.The proposed algorithm can improve these shortcomings, but it is not ideal. Our explanation of the results is that the dataset IRIS is relatively simple and the proposed algorithm is an attempt to the original algorithm when realizing global optimization. Our suggestion is that if the dataset contains amount of duplicate data, please apply the proposed algorithm. Anyway, the proposed algorithm is an attempt when traditional ones can’t find the global optimization, so it will spend more time.

Index Terms— clustering, FCM;kernel methods, simulated annealing, SAKFCM

Chengwei Jiang, Xiaoyan Cui, Tongliang Li
Department of Automation, Beijing University of Posts and Telecommunications, CHINA
Kaijian Zhang
School of Mechanical and Electronic Control Engineering, Beijing Jiaotong University, CHINA

[Download]


Cite: Chengwei Jiang, Xiaoyan Cui, Kaijian Zhang, Tongliang Li, "Kernel FCM Based on Simulated Annealing Algorithm," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 110-115, Beijing, 25-27 June, 2017.