Graph attention-based collaborative filtering for user-specific recommender system using knowledge graph and deep neural networks

摘要

Collaborative filtering suffers from the issues of data sparsity and cold start. Due to which recommendation models that only rely on the user–item interaction graph are insufficient to model the latent relationship between complex interaction of users and items. Existing methods utilizing knowledge graphs for recommendation explicitly model the multi-hop neighbors of an entity while ignoring the relation-specific as well as user-specific information. Moreover, a collaborative signal is also crucial to be modeled explicitly besides knowledge graph information. In this work, a novel end-to-end recommendation scenario is presented which jointly learns the collaborative signal and knowledge graph context. The knowledge graph is utilized to provide supplementary information in the recommendation scenario. To have personalized recommendation for each user, user-specific attention mechanism is also utilized. The user and item triple sets are constructed which are then propagated in the knowledge graph to enrich their representation. Extensive experiments are carried out on three benchmark datasets to show the effectiveness of the proposed framework. Empirical results show that the proposed model performs better than the state-of-the-art KG-based recommendation models.