Graph Neural Network-Enhanced Multivariate Time Series Forecasting with Series-Core Fusion

This study introduces Starformer, a hybrid model combining Graph Neural Networks (GNNs) with a novel Series-Core Fusion (SC-Fusion) mechanism for urban traffic prediction. By leveraging GNNs for spatial modeling and SC-Fusion for efficient temporal dependency capture, the model effectively addresses complex spatio-temporal dynamics in traffic systems. Evaluated on six widely used traffic datasets-METR-LA, PEMS-BAY, PEMS03, PEMS04, PEMS07, and PEMS08-Starformer demonstrates consistent and robust performance across diverse traffic conditions and regions. The results highlight its ability to model both short-term and long-term dependencies, making it well-suited for real-world applications. These findings emphasize the potential of integrating advanced neural network architectures for intelligent traffic management, contributing to smarter, more sustainable urban transportation systems.