Abstract: In order to solve the problems related to user’s privacies and single forms in detecting mail communities,the paper proposes an effective mail community partition method based on clustering ensemble.In a social network,the proposed method first utilizes the network structure features and mail attributes to measure the distances between nodes,instead of analyzing mail content that might be involved in user’s privacies.Then,the method generates initial clusters by the K-Means algorithm and adopts the co-association matrix to decide which cluster each node belongs to.Finally,according to the similarity between nodes in the co-association matrix,initial clusters are effectively combined to produce final communities.Experimental results show that the communities obtained by this algorithm are of high quality,although it does not use E-mail content,also it can find communities with various forms.