An Improved Density Peaks Clustering based on Rough Set Theory for Overlapping Community Detection

Mining community structure from network data set is an important research task in machine learning. Overlapping community detection is more complex due to the ambiguous of nodes which may be partitioned to different communities simultaneously. In this paper, an improved density peaks clustering is proposed to overlapping community detection. The rough set theory based uncertain similarity between nodes is defined in dual-nucleus subspace by fully considering the topological structure. Different strategies are used in density peaks clustering to improve the efficiency and the performance of the community division. Furthermore, rough set theory is employed to describe the overlapping nodes and rough set theory based overlapping community detection algorithm is proposed. Experiments are carried out on real-world social networks and artificial networks. The experimental results show that RSDPCD is effective.