Enhancing Temporal Link Prediction with HierTKG: A Hierarchical Temporal Knowledge Graph Framework
Mariam Almutairi, Melike Yildiz Aktas, Nawar Wali, Shutonu Mitra, Dawei Zhou
TL;DR
This work addresses the challenge of modeling rumor propagation as a temporal link-prediction task on evolving knowledge graphs. It introduces HierTKG, a framework that fuses Temporal Graph Networks with DiffPool-based hierarchical pooling to capture multi-scale temporal and structural dynamics, producing fused embeddings for prediction. Empirical results show state-of-the-art performance on structured datasets like ICEWS14, ICEWS18, and WikiData, with competitive results on noisy social-media data such as PHEME, and ablations highlight the importance of memory and attention in the fusion. The approach offers a scalable, robust tool for real-time rumor analysis and intervention by identifying key propagation moments and multi-level interaction patterns.
Abstract
The rapid spread of misinformation on social media, especially during crises, challenges public decision-making. To address this, we propose HierTKG, a framework combining Temporal Graph Networks (TGN) and hierarchical pooling (DiffPool) to model rumor dynamics across temporal and structural scales. HierTKG captures key propagation phases, enabling improved temporal link prediction and actionable insights for misinformation control. Experiments demonstrate its effectiveness, achieving an MRR of 0.9845 on ICEWS14 and 0.9312 on WikiData, with competitive performance on noisy datasets like PHEME (MRR: 0.8802). By modeling structured event sequences and dynamic social interactions, HierTKG adapts to diverse propagation patterns, offering a scalable and robust solution for real-time analysis and prediction of rumor spread, aiding proactive intervention strategies.