Community Detection Based on Joint Representation of Multi-Granular Networks

Community detection is an important problem in social network analysis. Most of the existing research on this topic is mainly based on single network. However, a single network cannot fully reflect the entire social relationship of an individual because of the diversity of social networks. To discover the community structures from multiple networks, a community detection algorithm based on the joint representation of multi-granular networks is proposed in this paper. First, the nodes in each network are embedded to obtain the corresponding vectors. Second, a joint representation of multi-granular networks is formed after the anchor nodes in each network interact to complement their information. Finally, an improved density peak algorithm called Center Density Peak algorithm (CDP) is proposed. Experiments on synthetic and real-world datasets show that the rich structural information of multi-granular networks can improve the accuracy of community detection.