Detecting network communities based on central node selection and expansion

Community detection aims to uncover the structure of closely connected nodes in complex networks, with applications in various fields such as social networks and biological networks. However, obtaining global information from a network remains a challenging task. Consequently, the study of local community detection has garnered widespread attention. Many existing algorithms for local community detection begin with selecting central nodes as initial communities and then expanding from there. However, the performance of community detection heavily relies on the selection of central nodes, the node updating order, and the community expansion strategy. To address these challenges, this study proposes an enhanced method based on central node selection and expansion (CNSE). Regarding the selection of central nodes, this study employs a voting approach involving three centrality methods, which comprehensively consider different centrality measures to choose the nodes with higher hybrid centrality as central nodes. For the node updating order, this study prioritizes the more important nodes to expedite their convergence process. During the community expansion, label vectors are propagated. This study considers both the similarity of neighbors and the influence of central nodes at different distances. Finally, key nodes are reassessed using community affiliation to ensure the accuracy of community detection. Experimental results on both real-world and synthetic networks demonstrate that CNSE has better performance in terms of normalized mutual information (NMI) and adjusted rand index (ARI).