Local Edge Betweenness Based Label Propagation for Community Detection in Complex Networks

Abstract

Nowadays, identification and detection community structures in complex networks is an important factor in extracting useful information from networks. Label Propagation Algorithm (LPA) with near linear-time complexity is one of the most popular methods for detecting community structures, yet its uncertainty and randomness is a defective factor. Merging LPA with other community detection metrics would improve its accuracy and reduce instability of LPA. Considering this point, in this paper we tried to use edge-betweenness centrality to improve LPA performance. On the other hand, calculating edge-betweenness centrality is expensive, so as an alternative metric, we use local edge-betweenness and present LPA-LEB (Label Propagation Algorithm Local Edge-Betweenness). Experimental results on both real-world and synthetic networks show that LPA-LEB possesses higher accuracy and stability than LPA when detecting community structures in networks.