Enhancing Recommender Systems Performance using Knowledge Graph Embedding with Graph Neural Networks

Recommender systems are a crucial component in personalized information retrieval, with their performance directly impacting user experience. However, traditional recommender systems often suffer from limitations in handling complex user-item interactions. Recently, knowledge graphs have emerged as valuable resources for enhancing recommendation effectiveness by capturing and utilizing a wide range of information. In this work, we propose a novel Auto-encoder-based structure integrated with Graph Neural Networks (GNN) layers for Knowledge Graph Embedding (KGE) tasks. This model effectively captures and leverages the rich information in knowledge graphs. Further, we combine this advanced GNN-based knowledge graph embedding model and a preference propagation mechanism inspired by RippleNet, to introduce a novel model, GraphRipple, aiming to enhance the recommendation performance. Extensive experiments on multiple datasets demonstrate the effectiveness of GraphRipple in improving recommendation quality.