Abstract: Community detection algorithms are the basic tools for discovering the internal structure and organizational principles of a community. Existing model-based and optimization-based algorithms usually consider two sources of information: network structure and node attributes, to obtain communities with both denser network structure and similar attribute information. However, such algorithms cannot automatically determine the relative importance between them and reveal subspaces, so the quality of the detected communities needs to be improved. This paper integrates subspaces into a new overlapping community detection framework, and designs an adaptive structure and attribute weighting strategy, which effectively reveals the subspace to discover diverse communities. Extensive experiments are conducted on artificial and real networks. The experimental results reveal the importance of subspaces on capturing better communities, justifying the rationality and effectiveness of the proposed algorithm.

Key words: view, overlapping, subspace, clustering, community detection