Hybrid-Aware Attention Network Incorporating Knowledge Graph for Recommendation

In recent years, knowledge graphs have been widely applied in the research of recommendation algorithms. However, most recommendation models focus on modeling item features in the relational space of the knowledge graph while neglecting the differences between the user side and the item side during feature extraction. Furthermore, they fail to comprehensively capture and fully utilize the high-order interaction relationships between users and items. To address these issues, we propose a method that combines knowledge graph and graph convolutional network, named the Hybrid-Aware Attention Network Incorporating Knowledge Graph for Recommendation (HAN). HAN jointly extracts the interaction information between users and items and the associative information of the knowledge graph to enable accurate recommendations. To validate the feasibility and effectiveness of the proposed model, experiments were conducted on three real datasets, demonstrating that the proposed algorithm achieves favorable recommendation performance compared to other benchmark models.