Time-Constrained Graph Pattern Matching in a Large Temporal Graph

Graph pattern matching (GPM) is an important operation on graph computation. Most existing work assumes that query graph or data graph is static, which is contrary to the fact that graphs in real life are intrinsically dynamic. Therefore, in this paper, we propose a new problem of Time-Constrained Graph Pattern Matching (TCGPM) in a large temporal graph. Different from traditional work, our work deals with temporal graphs rather than a series of snapshots. Besides, the query graph in TCGPM contains two types of time constraints which are helpful for finding more useful subgraphs. To address the problem of TCGPM, a baseline method and an improved method are proposed. Besides, to further improve the efficiency, two pruning rules are proposed. The improved method runs several orders of magnitude faster than the baseline method. The effectiveness of TCGPM is several orders of magnitude better than that of GPM. Extensive experiments on three real and semi-real datasets demonstrate high performance of our proposed methods.