UTCS: Effective Unsupervised Temporal Community Search with Pre-training of Temporal Dynamics and Subgraph Knowledge

2025-06-03

AI Summary

Existing community search methods on temporal graphs suffer from two key limitations: (i) reliance on predefined structural assumptions and (ii) inability to effectively model temporal dynamics. Method: This paper proposes the first unsupervised two-stage framework. In the offline stage, it employs self-supervised pretraining combined with multi-objective contrastive learning to jointly capture temporal interaction patterns and subgraph topology. In the online stage, it performs end-to-end, efficient community search using pretrained node representations and a differentiable community scoring function—eliminating the need for structural priors. Contribution/Results: By integrating temporal graph neural networks with a novel scoring mechanism, the method achieves an average 12.7% improvement in F1-score over state-of-the-art approaches across five real-world datasets, demonstrating significant gains in both accuracy and generalizability for dynamic community discovery.

Technology Category

• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Planning, Routing, and Scheduling: Temporal Planning
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Topic discovery and tracking

Abstract

In many real-world applications, the evolving relationships between entities can be modeled as temporal graphs, where each edge has a timestamp representing the interaction time. As a fundamental problem in graph analysis, {it community search (CS)} in temporal graphs has received growing attention but exhibits two major limitations: (1) Traditional methods typically require predefined subgraph structures, which are not always known in advance. (2) Learning-based methods struggle to capture temporal interaction information. To fill this research gap, in this paper, we propose an effective extbf{U}nsupervised extbf{T}emporal extbf{C}ommunity extbf{S}earch with pre-training of temporal dynamics and subgraph knowledge model ( extbf{model}). model~contains two key stages: offline pre-training and online search. In the first stage, we introduce multiple learning objectives to facilitate the pre-training process in the unsupervised learning setting. In the second stage, we identify a candidate subgraph and compute community scores using the pre-trained node representations and a novel scoring mechanism to determine the final community members. Experiments on five real-world datasets demonstrate the effectiveness.

Problem

• Traditional methods need predefined subgraph structures
• Learning-based methods fail to capture temporal interactions
• Proposes unsupervised temporal community search with pre-training

Innovation

• Unsupervised temporal community search model
• Pre-training temporal dynamics and subgraph knowledge
• Novel scoring mechanism for community detection