In order to solve the problem of privacy of group content in the process of overlapping social network partitioning, a social network partitioning method was proposed in this paper based on clustering and minimum spanning tree. This method which measures the distance of the node based on the close degree between attribute members has the inherent characteristics of social network, where the Kmeans algorithm is used to generate initial clustering and the clustering results could be expressed as matrix and band graph which use the Prim algorithm to get the minimum spanning tree and cut into K minimum spanning tree distance K edges of minimum spanning tree. And then, it could use the similarity of belonging to solve the outliers and the overlapping points of social network members. Experiments show that the algorithm does not involve the privacy of members of social networks. The community structure is of high quality, and the partitioning of overlapping members and isolated members could be effectively considered.