Highly efficient epidemic spreading model based LPA threshold community detection method

Community structure is the most basic and important topological characteristic of complex network and community detection method is therefore significant to its character statistics. Community detection in heterogeneous structured network is an attractive research problem while most of the previous approaches and algorithms attempt to divide networks into communities based on node or edge measurement. The Label Propagation Algorithm (LPA) adopted semi-supervised machine learning and implemented community detection intelligently with automatic convergent process of network entity label iteration. But LPA often resulted in low efficiency in the final convergent process. In this paper, aiming to promote the low running efficiency and converging rate of LPA in overlapping community detection and focusing on user label selection for behavior analysis, we proposed a new method ESLPA (Epidemic Spreading based LPA) for community detection using epidemic spreading model combined with network connection matrix's largest Eigenvalue as label propagation threshold. Extensive experiments in artificial network dataset and real-life large networks derived from online social media have been conducted to explore the optimal mechanism of the most suitable community detection method by virus infection threshold. According to experimental result, it has been proved that our method is more accurate and faster than several traditional modularity based community detection methods.