The item-based collaborative filtering recommendation algorithm and the resource diffusion algorithm based on the tripartite graph have some problems, such as low information utilization rate and sparseness of tag data. In order to solve these problems, an improved community recommendation model based on overlapping tags is proposed, which aims to make full use of the existing information and tags to personalize recommendation for users, shorten the time for users to find resources, and improve the quality of recommendation. Taking movie recommendation as an example, the proposed model transforms movie related information into tags, constructs complete tripartite graph through the relationship among users, movies and tags, introduces resource diffusion algorithm and K-means algorithm to find and divide overlapping communities of tags, and then makes personalized recommendation for users based on the overlapping communities of tags. Experiments on the hetrec 2011-movies-2k dataset show that the accuracy and recall rate of the proposed recommendation model are improved compared with the item-based collaborative filtering recommendation algorithm and the resource diffusion algorithm based on tripartite graph.