A Link Strength Based Label Propagation Algorithm for Community Detection

The Label Propagation Algorithm (LPA) is a fast algorithm for community detection. This algorithm has proved its efficiency and scalability even in large social networks. It is basedonly on updating the label of each node by the most frequent label within its neighbors. However, in case of many maximal labels, a random choice is made. This reduces the performance of the algorithm. In this paper, we propose a new version of LPA, called Link Strength-based LPA (LS-LPA), that addresses this issue. In our approach, we rank labels according to the link strength of each neighbor. The strength of links is quantified via k-neighborhood which includes common neighbors and multi-step neighbors. Experiments on different networks with different sizes show that our solution improves LPA's performance significantly.