Reliable Community Search over Dynamic Bipartite Graphs

Mo Li, Zhiran Xie, Linlin Ding

Abstract: Bipartite graphs naturally represent relationships between two different entities, such as people-location networks, author-paper networks. In dynamic scenarios like product recommendation systems, where purchase behavior and interests evolve, dynamic bipartite graphs are emerged. Despite extensive research on community search in (dynamic) unipartite graphs, dynamic bipartite graphs remain unexplored. While the duration of a community reflects its temporal continuity, managing its size is also crucial. Over time, a community’s prolonged existence tends to decrease in size, posing a practical challenge. To fill this research gap, we introduce a reliable community model over dynamic bipartite graphs that accounts for time span, size and degree constraints. Then, we propose an efficient RCSearch algorithm to solve the reliable community search, leveraging properties of reliable \((\alpha , \beta )\)-communities and dynamic programming strategies. Furthermore, effective optimization strategies are devised to accelerate this process. Finally, extensive experiments conducted over 7 real-world graphs demonstrate the effectiveness and efficiency of our proposed methods.