Abstract: In the process of minimization Fuzzy C-Means (FCM) clustering objective function, to solve the problem of lacking theoretical foundation and effective evaluation methodology in determining fuzzy weighted exponent "m" at present, a fuzzy weighted exponent algorithm based on fuzzy relevance was put forward. Firstly, valid function was defined based on Fuzzy relevance, then the validity of FCM clustering was calculated by Gauss iteration and its result was returned to the change of fuzzy weighted exponent, the fuzzy weighted exponent "m" will be converged to a stable optimum resolution. This algorithm is proved to be effective by theoretical analysis and experiments, and the weighted exponent "m" got from this algorithm conforms to prospective result.

Key words: fuzzy weighting exponent, Fuzzy C-Means （FCM）, clustering validity, fuzzy relevance

摘要： 在极小化模糊C均值（FCM）聚类目标函数的过程中，针对目前模糊加权指数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Gauss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。

关键词: 模糊加权指数, 模糊C均值, 聚类有效性, 模糊相关度