Dataset Statistics.

Recommender systems have recently gained significant traction as powerful tools for personalized content delivery. While accuracy remains a key focus, users now expect more than precise suggestions. To meet diverse preferences, these systems must also ensure recommendation diversity. They are widely applied in domains such as e-commerce, social media, and online entertainment platforms. Conventional approaches like collaborative filtering mainly emphasize user–item interactions, often overlooking contextual and attribute information, which results in limited performance, especially under sparse data conditions. To address this, we present the KGRec- Knowledge Graph Attention Network Recommendation model, the novel KGRec model integrates knowledge graphs to capture higher-order relationships among users, items, and their associated attributes. KGRec applies multi-layer embedding propagation combined with an attention mechanism to model indirect user–item connections through intermediate attributes, enabling the model to assess the significance of each relation and thus improve recommendation quality. Empirical evaluations on four benchmark datasets— Yelp2018, Last-FM, Amazon-Book and MovieLen-1M—demonstrate the effectiveness of KGRec. The proposed model consistently outperforms all baseline methods across every evaluation metric. These improvements highlight the model’s robustness and its effectiveness in capturing richer semantic representations for recommendation.