Epdemic spreading model based overlapping community detection

Community detection in inhomogeneous structured network is an attractive research problem that searches for methods to discover groups in which individuals are more densely interconnected with each other with higher probability of internal information propagation. While most of the previous approaches attempt to divide networks into communities according to the algorithm results of network or edge measurement, Label Propagation Algorithm (LPA) adopts semi-supervised machine learning and implements community detection in an intelligent way with the automatic convergent process of network entity label iteration. In this work, we study the early community detection approaches, explore the low efficacy and stagnant converging rate of LPA in its response to network with overlapped communities, and propose a new approach for community detection using epidemic spreading virus to discover groups with super positioned members. Extensive experiments in synthetic signed network and real-life large networks derived from Internet social media are conducted to explore the optimal mechanism of the most suitable community-detecting virus infection.