Empowering Spatiotemporal Traffic Prediction Through Fusing Flowformer and GNN

This study introduces STFLformer, a hybrid model that integrates GNNs with Flowformer for urban traffic prediction. By leveraging GNNs to model spatial dependencies and Flowformer for capturing temporal patterns, STFLformer effectively addresses complex spatio-temporal dynamics in traffic systems. Extensive experiments conducted on six benchmark datasets, including METR-LA, PEMS-BAY, PEMS03, PEMS04, PEMS07, and PEMS08, demonstrate the model's superior performance across diverse traffic scenarios. Notably, STFLformer consistently outperforms state-of-the-art models, highlighting its robustness and adaptability to real-world traffic patterns. These findings underscore the potential of advanced neural architectures in fostering intelligent traffic management, promoting sustainable urban mobility, and contributing to improved air quality.