مشخصات پژوهش

Community detection is still one of the interesting and hot topics in the field of complex networks. Among the community detection algorithms, Label Propagation Algorithm (LPA), due to its nearly linear time complexity and fast execution of the algorithm, has attracted a lot of attention. But random selection of nodes and the method of updating of labels has turned this algorithm into an unreliable algorithm without stable results. In this paper a new multi factor node scoring-based label propagation algorithm (MFNS-LPA) is proposed. Four different factors are adopted to evaluate the importance of nodes which are: nodes similarity, nodes degree, K-shell value, and the percentage of the important neighbors of a node. Besides proposing a new measure for scoring nodes, the label updating strategies are improved so that the convergence speed of the algorithm is significantly decreased. Experiments are performed on real-world and synthetic networks to evaluate the performance of different methods. Results indicate that the proposed algorithm outperforms other methods in terms of accuracy, convergence, and fast execution.