A Time-Aware Recommendation Model with Social Impact

Sequential recommendation has emerged as a significant research focus within the realm of recommender systems. This approach often involves modeling user interactions with items as isomorphic graphs, where nodes represent items and edges denote interactions. Graph Neural Networks (GNNs) have been effectively employed to capture higher-order relationships between nodes in these graphs, enhancing the quality of recommendations. Despite the success of graph-based models in sequential recommendation, there is a notable gap in the literature regarding the integration of social influence. Many existing studies concentrate primarily on item interactions without adequately addressing how social factors might affect sequential recommendations. To address this gap, we propose the Time-Aware Recommendation Model with Social Influence (TARMSI). This model emphasizes the role of social relationships in sequential recommendation by incorporating social influence into the recommendation process. TARMSI operates on a fine-grained level, considering both item categories and interaction times to enhance the recommendation quality. Specifically, it introduces social relations into the user representation framework, allowing for a more nuanced understanding of item interactions based on social contexts. In addition to integrating social influence, TARMSI employs temporal self-attention networks to model users’ dynamic preferences. This approach enables the model to adapt to changes in user preferences over time, further improving the accuracy of recommendations. The effectiveness of the TARMSI model is demonstrated through extensive experiments conducted on three mainstream datasets, showcasing its superiority over existing methods in capturing the influence of social interactions and temporal dynamics on sequential recommendations. By addressing the limitations of previous models, TARMSI offers a more comprehensive and contextually aware approach to recommendation systems.