In many applications, a cluster structure in a given dataset is often confined to a subset of features rather than the entire feature set. One of the main problems is how to make use of all the features effectively and adequately to discover structures. By using weighted dissimilarity measure and adding weight entropy regularization term to the objective function, a novel fuzzy c-means algorithm is developed for clustering and feature selection. It can automatically calculate the weights of all attributes in each cluster, and simultaneously minimizes the within cluster dispersion and maximizes the attribute weight entropy to stimulate attributes to contribute to the identification of clusters. Experiments on real world datasets show the effectiveness of this algorithm compared with other well known clustering algorithms. © 2011 IEEE.