KGRec: A knowledge graph attention-based model for recommender system.

Area of Science:

• Computer Science
• Artificial Intelligence
• Information Retrieval

Background:

• Recommender systems are crucial for personalized content delivery, but often lack diversity.
• Traditional methods like collaborative filtering struggle with sparse data and overlook contextual information.
• Enhanced recommendation quality requires capturing higher-order relationships beyond direct user-item interactions.

Purpose of the Study:

• To introduce KGRec, a novel Knowledge Graph Attention Network Recommendation model.
• To improve recommendation accuracy and diversity by integrating knowledge graphs.
• To address limitations of conventional recommender systems in handling sparse data and contextual information.

Main Methods:

• Developed KGRec, a model integrating knowledge graphs to capture user, item, and attribute relationships.
• Employed multi-layer embedding propagation and an attention mechanism to model indirect user-item connections.
• Utilized knowledge graphs to assess the significance of relational attributes for improved recommendation quality.

Main Results:

• KGRec consistently outperformed baseline methods across four benchmark datasets (Yelp2018, Last-FM, Amazon-Book, MovieLen-1M).
• The model demonstrated superior performance in recommendation accuracy and diversity.
• Empirical evaluations confirmed the effectiveness of KGRec in capturing richer semantic representations.

Conclusions:

• KGRec effectively leverages knowledge graphs to enhance recommender systems.
• The model's attention mechanism and embedding propagation capture complex relationships, improving recommendation quality.
• KGRec offers a robust solution for personalized content delivery, addressing limitations of traditional approaches.