A novel recommender system using light graph convolutional network and personalized knowledge-aware attention...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Graph Neural Networks (GNNs) are increasingly used in recommender systems (RS) for feature extraction and relationship modeling.
• GNNs face challenges in capturing fine-grained knowledge graph (KG) semantics and effectively modeling user-item interactions.
• Personalized knowledge-aware recommendation offers a promising approach to address these limitations.

Purpose of the Study:

• To propose a novel recommender system, LGKAT, that combines user-item graphs and knowledge graphs for improved recommendation accuracy.
• To enhance the modeling of user-item interactions by leveraging rich semantic information from KGs.
• To address the limitations of existing GNN-based recommender systems in capturing nuanced relationships.

Main Methods:

• Developed LGKAT, a recommender system integrating user-item and knowledge graphs.
• Employed Light Graph Convolutional Network (LightGCN) for efficient management of user and item embeddings.
• Introduced an attention sub-network to encode KG semantics into personalized item embeddings.

Main Results:

• Extensive experiments were conducted on four benchmark datasets.
• LGKAT demonstrated significant superiority over state-of-the-art methods in terms of F1_score and recall.
• The integration of LightGCN and attention mechanisms effectively improved recommendation quality.

Conclusions:

• LGKAT effectively tackles limitations in current recommender systems by integrating knowledge graphs.
• The proposed method achieves superior performance in personalized knowledge-aware recommendation tasks.
• This research highlights the potential of combining GNNs with knowledge graphs for enhanced recommender systems.