基于模糊相关度的模糊C均值聚类加权指数研究.pdfVIP

第 30 卷第 12期 计算机应用 Vol. 30 No. 12 2010年 12月 Journal of ComputerApp li ations De . 2010 : 100 1- 9081( 20 10) 12- 3388- 03 C 肖满生, 阳娣兰, 张居武, 唐文评 ( , 412008) (x iaomansh eng@ tom. om ) : 在极小化模糊 C均值 ( FCM )聚类目标函数的过程中, 针 目前模糊加权指数m 的确定缺乏理论依据和 有效评价方法的问题, 提出了一种基于模糊相关度的模糊加权指数计算方法首先定义模糊相关度的聚类有效性函 数, 然后通过 G auss迭代计算 FCM 聚类有效性并将其反馈到模糊加权指数的变化中, 从而使m 收敛到一 个稳定的最 优解理论分析和实验结果表明, 该算法是有效的, 所得到加权指数m 符合预期的结果 : 模糊加权指数; 模糊 C 均值; 聚类有效性; 模糊相关度 : TP39 1 : A R esearch of w eigh ting exponent of fuzzy Cm ean s a lgorithm based on fuzzy relevance X IAO M ansheng, YANG D ilan, ZHANG Juw u, TANG W enp ing (Colleg e of S cience and Technology, H unan Un iversity of Technology, Zhuzh ouH unan 412008, China) A bstract: In the pro ess of m inmi ization Fuzzy CM eans ( FCM ) lustering obje tive fun tion, to solve the problem of la king theoreti al foundation and effe tive evaluation m ethodology in determ in ing fuzzy weighted exponent m at present, a fuzzy w eighted exponent algorithm based on fuzzy relevan e was put forward. F irstly, valid fun tion w as defined based on Fuzzy relevan e, then the valid ity ofFCM lusteringw as al ulated by Gauss iteration and its resu ltw as returned to the hange of fuzzy w eigh ted exponent, the fuzzy weighted exponen t m w ill be onverged to a stable optmi um resolution. Th is algorithm is proved to be effe tive by theoreti al analy sis and expermi ents, and the w eighted exponent m got from th is algorithm onform s to prospe tive result. K ey w ords: fuzzy w eighting exponen t; Fuzzy CM eans ( FCM ); lustering validity; fuzzy relevan e , m 0 引言