The Temporal Knowledge Graph Reasoning Model Based on Graph Attention Network

Temporal knowledge graphs play a crucial role in event prediction and reasoning tasks, but current techniques face challenges in terms of interpretability and capturing complex temporal patterns. To address these limitations, this study proposes the TGAT-Logic model, which combines a Temporal Graph Attention Network (TGAT) with temporal logic rules to enhance the model's understanding of complex temporal structures, reasoning capabilities, and interpretability. The model consists of four core components: Temporal Graph Attention Network, intelligent random walk sampler, temporal logic rule generator, and rule applier, which work collaboratively to significantly improve the accuracy of link prediction and reasoning tasks. Experimental results show that TGAT-Logic performs exceptionally well on several public datasets, validating its effectiveness in temporal knowledge graph reasoning tasks and providing new research perspectives for this domain.