With the rapid growth of users, traditional collaborative filtering methods continue to struggle with handling data sparsity and cold start issues, leading to significantly reduced recommendation accuracy. To address these persistent challenges, this study proposes a movie recommendation system that integrates Graph Neural Networks (GNNs) with temporal contextual information. GNNs model user-movie interactions as graph structures, with users and movies represented as nodes and their interactions as edges. By incorporating temporal context, the model captures dynamic user preferences that evolve over time, allowing for more personalized and context-aware recommendations. The GNN model achieved an RMSE of 1.51, which further improved to 1.45 with the inclusion of temporal context, demonstrating the crucial role of contextual information in enhancing the system’s ability to predict user behavior. These findings highlight the substantial potential of integrating GNN with contextual data to significantly improve the overall performance of recommendation systems, especially in scenarios characterized by sparse data or limited user-item interactions.
Read full abstract