An Overlapping Community Detection Algorithm based on Density Peaks and Label Propagation

Abstract

Many nodes in real networks belong to multiple communities at the same time. The research methods of overlapping community detection have attracted more and more attention. Density peak clustering (DPC) and label propagation algorithm (LPA) can efficiently complete community detection. However, DPC cannot distinguish overlapping nodes in the network, and the stability of LPA is very poor. In this paper, we propose an overlapping community detection algorithm based on density peaks and label propagation (ODLPA). ODLPA uses the degree of neighbor nodes to calculate the local density of nodes and expands the selection range of community centers. At the same time, ODLPA uses and improves the Speaker-listener rule to stably and efficiently detect overlapping communities in the initial community. ODLPA also proposed a community integration method based on greedy strategy to further improve the quality of community division. Finally, the effectiveness of the proposed algorithm is verified by comparison with other related algorithms on real-world network.