Abstract: Community dipartition is used in biochemistry, sociology, eco-systems, etc. The reliability of the results and the efficiency of the algorithm are the focus of the study. The Louvain algorithm is an algorithm with relatively reliable result and better efficiency. In this paper, the Louvain algorithm is improved in dealing with the leaf nodes. By studying the characteristics of the leaf nodes and the inadequacies of Louvain algorithm, the improved algorithm prunes the leaf nodes to avoid redundant computation. 18 sets of artificial data and the email data of our school are respectively processed using improved algorithm and Louvain algorithm. The comparison of results shows that the improved algorithm improves the processing speed while maintaining the result reliable.