StreamSC: A Learning-Based Framework for Efficient Subgraph Counting in Stream Graphs
Zhen Xie, Xiang Hui Zhao · 2025
Graphs serve as essential data structures for modeling intricate relationships among entities across various domains, including social networks and chemical reactions. When dealing with large-scale graph data, the stream graph approach is frequently employed to process and manage the data efficiently. The subgraph counting problem in stream graphs is recognized as a important and challenging task, primarily because its core component, subgraph matching, is classified as NP-complete. In this work, we propose deep learning-based solution framework named StreamSC to solve the subgraph counting in stream graphs. This framework offers two key advantages: (i) It's the first learning-based framework to address the subgraph counting problem focused on stream graphs; and(ii) this framework addresses the issue of dynamic changes in the topology of the data graph caused by the addition or deletion of edges in stream graphs through specialized design and optimization.