Title:Fast Algorithms for Intimate-Core Group Search in Weighted Graphs

Abstract:Community search that finds query-dependent communities has been studied on various kinds of graphs. As one instance of community search, intimate-core group search over a weighted graph is to find a connected $k$-core containing all query nodes with the smallest group weight. However, existing state-of-the-art methods start from the maximal $k$-core to refine an answer, which is practically inefficient for large networks. In this paper, we develop an efficient framework, called local exploration k-core search (LEKS), to find intimate-core groups in graphs. We propose a small-weighted spanning tree to connect query nodes, and then expand the tree level by level to a connected $k$-core, which is finally refined as an intimate-core group. We also design a protection mechanism for critical nodes to avoid the collapsed $k$-core. Extensive experiments on real-life networks validate the effectiveness and efficiency of our methods.