Multiresolution Local Spectral Attributed Community Search

Community search has become especially important in graph analysis task, which aims to identify latent members of a particular community from few given nodes. Most the existing efforts focus on exploring structure with single scale nodes are located. Despite promising results, following two insights often neglected. First, node attributes provide rich and highly related auxiliary information apart network interactions for characterizing properties. Attributes may indicate assignment very links, would be difficult determine alone. Second, multiresolution affords depict hierarchical relation ensure that one them is closest real one. It essential users understand underlying explore strong attribute cohesiveness at disparate scales. These aspects motivate us develop new framework Multiresolution Local Spectral Attributed Search (MLSACS). Specifically, inspired by local modularity, wavelets scaling functions, we propose modularity (MLQ) based reconstructed node-attribute graph. Furthermore, detect communities cohesive structures different scales, sparse indicator-vector developed MLQ solving linear programming problem. Extensive experimental results both synthetic real-world attributed graphs have demonstrated detected meaningful can changed reasonably.