摘要
文章提出了将HCM,FCM和核方法结合在一起的,一种改进模糊核聚类算法。该算法的思想是将样本数据映射到特征空间,然后在特征空间内计算类中心、隶属度以及距离表达式,再在特征空间内进行模糊聚类,并且针对个别样本(即隶属度比较接近的样本)加入了截集因子确定样本的归属,确保聚类的效果。实验结果表明,与传统的模糊聚类算法相比,改进的模糊核聚类算法在多种数据结构条件下可以有效地进行聚类,总体性能优于HCM,FCM和FKCM。
This paper presents a Sectional Set FKCM,which is a generalization of the conventional Sectional Set FCM and HCM.The main idea of the algorithm is how to map sample data to feature space and compute the center and degree of membership for all categories.So the fuzzy clustering can be processed in the feature space by adding cut factor(to some sample(process close degree of membership)to conform their adscription and ensure a good effect of clustering.The results gained by experiments using actual data show that the Sectional Set fuzzy kernel C-means clustering algorithm can effectively cluster for data with diversiform structures compared with the Sectional Set fuzzy C-means clustering algorithm.The algorithm possesses better performance than classical clustering algorithm(HCM,FCM,FKCM).
出处
《西华大学学报（自然科学版）》
CAS
2007年第3期48-50,69,共4页
Journal of Xihua University:Natural Science Edition
关键词
核方法
模糊C-均值
核聚类算法
特征空间
kernel methods
fuzzy C-means
kernel clustering algorithm
feature space
