：Since the general fuzzy ??-means (FCM) algorithm requires sum of membership satisfying the normalization condition for a sample to each cluster, and thus results algorithm sensitive to noise or outliers and reducing the validity of the clustering on non-equilibrium distribution samples. Therefore, an FCM clustering algorithm with the improved fuzzy membership constraint function is proposed. By relaxing the normalization condition, a new formula of membership division is deduced, and the membership is constantly corrected in the clustering process, so that it will eliminate the noise sample, and improve the valid