Efficient Subgraph Similarity All-Matching

Being a fundamental problem in managing graph data, subgraph exact all-matching enumerates all isomorphic matches of a query graph q in a large data graph G. The existing techniques focus on pruning non-promising data graph vertices against q. However, the reduction and sharing of intermediate matches have not received adequate attention. These two issues become more critical on subgraph similarity all-matching due to the (possibly) massive number of intermediate matches. This paper studies the problem of efficient subgraph similarity all-matching by developing a novel query processing framework. We propose to effectively decompose a query graph into a hierarchical structure with the aim to minimize the number of intermediate matches and share intermediate matches. Novel techniques are then developed to estimate the number of intermediate matches, efficiently merge the intermediate matches, and generate efficient query execution plans. Experimental on real and synthetic datasets show that our approach outperforms the state-of-the-art approach for orders of magnitude.