Local Spectral for Multiresolution Community Search in Attributed Graph

Qingqing Li, Huifang Ma, Zhixin Li, Liang Chang

Abstract: Community search aims to identify all latent members of a particular community from given nodes. Though previous studies have been proven effective, some of them ignore two insights. First, node attributes provide side information to describe features of nodes. It contributes to optimal results. Second, multiresolution community affords latent information to depict the hierarchical relation of network and ensure that one of them is closest to the real one. These aspects motivate us to propose LSMSA, which utilizes Local Spectral for Multiresolution community Search in Attributed graph. Specifically, inspired by the local modularity and graph wavelets, we propose Multiresolution Local modularity (MLQ) in node-attribute graph. Furthermore, to detect multiscale local communities, a sparse indicator-vector is developed based on MLQ via solving a linear programming problem. Extensive experimental results on real-world attributed graphs have demonstrated the detected community is meaningful and its scale can be changed reasonably.