A two-stage BFS local community detection algorithm based on node transfer similarity and Local Clustering Coefficient

Local community detection is of great value to network analysis, because it can efficiently find the community in which the given node located. However, the seed-dependent, core-criteria and termination problems are still difficulties for the major detection frames, which explore the core community and extend the community based on evaluation functions. In this paper, we proposed a local community detection algorithm (TSB) based on Breadth First Search (BFS), the proposed node transfer similarity and Local Clustering Coefficient (LCC). In our modeling, the initial clustering ability is strengthened by clustering the direct neighbor node with high clustering coefficient, so the seed-dependent problem could be avoided. We gave a new core criteria to evaluate the candidate core nodes, which is combined of LCC, the node transfer similarity and BFS depth. The dynamic termination thresholds for each candidate node in the core and extension stages are separately relying on its father node and the average node transfer similarity. In the experimental work, we compared our method with several previous methods through real-world and artificial synthetic networks. The results show that our method has better performance on accuracy with various community structure.