CSMN: An Efficient Continuous Subgraph Matching Network for Dynamic Graphs

Subgraph matching (or subgraph isomorphism) aims to find all isomorphic embeddings of a query graph in a data graph, while in real scenarios, the graph data such as social networks are updated frequently. To deal with the subgraph matching problem in dynamic graphs, continuous subgraph matching has been proposed. However, methods of solving continuous subgraph matching problems are still time-consuming due to their logical operations. In recent years, graph neural networks (GNNs) have achieved great success in capturing graph information. Therefore, several GNN-based methods have been proposed to solve the subgraph matching problem. Although GNN-based methods are more efficient than traditional methods, they only can find the matching results on the static data graph, and there are no methods dedicated to the problem of continuous subgraph matching. To fill this gap, in this paper, we propose an efficient Continuous Subgraph Matching Network (CSMN). CSMN can efficiently construct embeddings of vertices in dynamic graphs and improve the efficiency of the continuous subgraph matching process with Time-aware Graph Attention Network (TGAN) and Time-informer Attention Network (TIAN). We conducted our experiment in three real-world datasets. The results demonstrate that CSMN can not only improve the time efficiency but also attain enhancements in accuracy.