Hybrid-order Stochastic Block Model

Community detection is a research hotspot in machine learning and data mining. However, most of the existing community methods only rely on lower-order connectivity patterns, while ignoring higher-order unable to capture building blocks complex network. In recent years, some based structures have been developed, but they mainly focus motif network composed structures, which violate original topological structure are affected by fragmentation issue, resulting deviation results. Therefore, there still lack that can effectively utilize patterns patterns. To overcome above limitations, this paper proposes Hybrid-order Stochastic Block Model (HSBM) from perspective generative model. Based classical stochastic block model, generation modeled uniformly, properties maintained using At same time, heuristic algorithm for proposed optimize objective function. Extensive experiments six real-world datasets show method outperforms approaches.