LDA-LPA: A Community Detection Method Based on Topic Model

Community detection in complex networks is an important task in network analysis. And community structures in networks can be detected from many aspects. However, community detection based on topic model is rarely studied, and most of the existing community detection methods based on topic model are generally based on node or edge attributes, such as text information. In this paper, a novel community detection method based on topic model is proposed, which only requires the structural information of networks, and does not need node or edge attributes to be presented. In addition, this paper also proposes a new graph embedding method based on the neighborhoods of node, which is simpler and does not need the complicated training process. Experimental results show that the community detection method proposed in this paper is superior to most benchmark algorithms for community detection. In particular, on some datasets, the method proposed in this paper can achieve the division of normalized mutual information of 1.0, that is, completely correct community division results can be obtained.