CK-LPA: Efficient community detection algorithm based on label propagation with community kernel

Abstract

With the rapid development of Web 2.0 and the rise of online social networks, finding community structures from user data has become a hot topic in network analysis. Although research achievements are numerous at present, most of these achievements cannot be adopted in large-scale social networks because of heavy computation. Previous studies have shown that label propagation is an efficient means to detect communities in social networks and is easy to implement; however, some drawbacks, such as low accuracy, high randomness, and the formation of a “monster” community, have been found.In this study, we propose an efficient community detection method based on the label propagation algorithm (LPA) with community kernel (CK-LPA). We assign a corresponding weight to each node according to node importance in the whole network and update node labels in sequence based on weight. Then, we discuss the composition of weights, the label updating strategy, the label propagation strategy, and the convergence conditions. Compared with the primitive LPA, existing drawbacks are solved by CK-LPA. Experiments and benchmarks reveal that our proposed method sustains nearly linear time complexity and exhibits significant improvements in the quality aspect of static community detection. Hence, the algorithm can be applied in large-scale social networks.