Gated Recurrent Uni-Spatiotemporal Graph Network Recommendation Algorithm for Heterogeneous Travel Data

With the widespread application of location-based social networks in personalized recommendation, accurately capturing user behavior and social relationships has become a key research focus. This paper proposes a GAT-GRU-based spatiotemporal aggregation POI recommendation framework, designed to enhance POI recommendation accuracy. By leveraging a Graph Attention Network to model user social relationships and POI regional factors, and integrating GRU with a spatiotemporal aggregation function to refine user behavior representations, the model achieves multidimensional feature fusion and spatiotemporal awareness. Experimental results demonstrate that UPGAT-TSAG outperforms existing baseline models across multiple real-world datasets, exhibiting superior recommendation accuracy and robustness in modeling complex social and spatial relationships.