Community Division Algorithm Based on Node Similarity and Multi-attribute Fusion

In order to improve the accuracy of community division results, a community division algorithm (NSMF) based on node similarity and multi-attribute fusion is proposed. NSMF algorithm iteratively selects the node with the highest PageRank value as the initial clustering center of the community by improving the PageRank algorithm. The initial clustering node is selected through the network global information, which effectively avoids the low stability of the random selection of the initial node. Then, the node with the highest similarity with the community node is added to the community. At this time, only the local information of the network is calculated, the computational strength of the network is reduced. This paper compares with GN algorithm and FG algorithm on three real network data sets with different community structures. The results show that NSMF algorithm has the best overall performance in correct partition rate, global modularity and standardized mutual information.