CGQ: Relationship-Aware Contextual Graph Querying in Large Networks.
The phenomenal growth of graph data from a wide-variety of real-world applications has rendered graph querying to be a problem of paramount importance. Traditional techniques use structural as well as node similarities to find matches of a given query graph in a (large) target graph. However, almost all previous research has tacitly ignored the presence of relationships and context (usually manifested in the form of node/edge label distributions) in the query. In this paper, we propose CGQ -- Relationship-Aware Contextual Graph Querying} for real-world graphs. Given a query graph and a target graph, CGQ identifies the (top-k) maximal common subgraphs between the query and the target graphs with the highest contextual similarity. We prove that the problem is NP-hard and APX-Hard. To overcome this computational bottleneck, we propose a hierarchical index, CGQ-Tree, with its associated CGQ search algorithm. Empirically, the CGQ search algorithm is capable of achieving speed-ups of up to three orders of magnitude over a baseline strategy. Our experiments show that CGQ is effective, efficient and scalable.
Publisher URL: http://arxiv.org/abs/1801.06402