Recommendation System Based on Temporal Knowledge Graph Path Reasoning
Haoyuan Ren, Liangzhong Cui · 2023
Reasoning based on Knowledge graph has been studied and used for explanation recommendations because it can provide a clear explanation. However, the current recommendation methods based on Knowledge graphs use static Knowledge graphs, without considering the time information that has a significant impact on recommendation. In this paper, we propose a method of path recommendation reasoning on the temporal Knowledge graph. This method uses the time information of user and item interaction to provide better recommendations with more reasonable explanations. First, we use a representation learning method that has a good performance effect on the temporal Knowledge graph to code entities and relations, and then use the path reasoning method based on Reinforcement learning to recommend according to the embedded representation of entities and relationships. We conducted extensive experiments on three real-world datasets. The results indicate that the proposed method is effective in NDCG@K Recall@K Precision@K and HR@K Four evaluation indicators are superior to existing models.