|
چکیده
|
Recommendation systems play a crucial role in helping users navigate the vast amount of information available online, enhancing user experience and engagement. With the increasing complexity of user-item interactions, leveraging advanced techniques such as Graph Neural Networks (GNNs) has become essential for improving the accuracy and relevance of recommendations. GNNs effectively model the relationships between users and items, capturing intricate patterns in user behavior and preferences. In this context, we propose the Position-Aware Neural Graph Collaborative Filtering (PA-NGCF) model, which introduces two key innovations that distinguish it from existing graph neural network-based recommender systems. Our primary contribution is the development of a novel method for creating node embeddings that utilize limited rating information from users to effectively capture user-item interactions. Additionally, our model explicitly incorporates the positional information of nodes within the graph structure during the message passing phase, allowing for a more nuanced understanding of the role and importance of each node in the recommendation process. Extensive experiments conducted on two real-world datasets demonstrate the effectiveness of our approach. The results indicate that the PA-NGCF model significantly outperforms existing baselines in terms of accuracy, showcasing the advantages of integrating positional awareness and higher-order neighborhoods in graph-based recommendation systems. Our findings highlight the potential of GNNs in enhancing the quality of recommendations and addressing challenges such as cold start problems in sparse datasets.
|