Personalized Recommendation System of E-Commerce in the Digital Economy Era: Enhancing Social Connections with Graph Attention Networks

In the era of digital economy, e-commerce platforms face significant challenges in providing personalized recommendations due to data sparsity, a result of limited user-product interactions. This research introduces a model that enhances e-commerce personalization by leveraging graph attention networks to mine and integrate information from user-item (U-I) score graphs and user-user (U-U) social graphs. The U-U social graph captures user relationships, providing supplementary information to uncover social connections and preferences, especially among users with sparse interactions, thereby alleviating data sparsity concerns. Additionally, the model incorporates graph contrastive learning to extract universal features of users and items, further mitigating the sparse data challenge. Our approach, when applied to the CiaoDVD and Epinions datasets, demonstrated a 1.21% improvement in Mean Absolute Error (MAE) and a 2.22% improvement in Root Mean Square Error (RMSE) on the CiaoDVD dataset, and a 1.71% improvement in MAE and a 2.11% improvement in RMSE on the Epinions dataset, outperforming all baseline methods.