Local Fuzzy Community Detection in Networks

Background: Community detection is significant for the understanding of the structure and function of networks, and becomes an attractive topic for researchers. However, many existing local methods only focus on disjoint communities and some recently proposed overlapping community detection methods are global methods with high computational cost. Objective: To improve the accuracy and speed of community detection and obtain the fuzzy coefficients of overlapping nodes with low computational cost, a local fuzzy agglomerative method is proposed in this paper. Methods: In the detection process, each local community is determined based on community strength. The overlapping communities and fuzzy coefficients of nodes are obtained by coordinating and normalizing the contribution of the overlapping nodes to their belonging communities. Results: Theoretical analysis and data simulations show that our local method can detect disjoint and overlapping communities in linear time with the network size. The overlapping communities and the fuzzy coefficients of overlapping nodes are obtained accurately. Conclusion: The accuracy of our method is higher than the existing local methods for detecting disjoint communities. And it also detects the overlapping communities same as the global overlapping methods but with remarkably low computational cost.