Parallel overlapping community detection algorithm in complex networks based on label propagation

Some of the defects of community detection algorithms in complex networks include pre-parameter limits and redundant labels. Accordingly, we present an improved overlapping community detection algorithm in complex networks that is a minimal maximal clique label propagation algorithm (MMCLPA). The algorithm uses the minimal maximal clique (MMC), and allows each MMC to share the same label by reducing redundant labels and random factors. Therefore, the MMCLPA algorithm increases the stability of the algorithm. Mean-while, the MMCLPA algorithm completes label propagation from core node groups (MMC) for distribution using intimacy as the weight of the label. For the labels post processing, uses adaptive threshold method to overcome pre-parameter limitations in unknown complex networks. When compared with other community detection al-gorithms using arti cial and real network data, our experiment results prove that the MMCLPA algorithm not only increases the tolerance for mixing parameters, but also improves the robustness of the algorithm. Lastly, by employing a distributed computing model (Map Reduce) and cloud platform (Hadoop), we allow parallelization implementation of MMCLPA. Our experiment results show that parallel MMCLPA owns an approximation of quality compared with the stand-alone system, as well as better scalability.