Wavelet-Enhanced Spatio-Temporal Graph Attention Network with Weather Fusion for Traffic Flow Prediction

In recent years, with the acceleration of the urbanization process, the problem of traffic congestion has become increasingly severe. Accurate traffic flow prediction is crucial for urban traffic management and intelligent transportation systems. Aiming at the deficiencies of existing traffic flow prediction models in multi-scale feature decoupling and the spatiotemporal heterogeneity modeling of meteorological factors, this paper proposes a Spatio-Temporal Graph Attention Network (WSTAN-WF) based on the Maximal Overlap Discrete Wavelet Transform (MODWT) and a cross-modal gated attention mechanism. The model decomposes the traffic flow into multi-frequency band features through MODWT and uses frequency-divided temporal attention to achieve adaptive fusion. The designed gated module can dynamically model the spatial impact of meteorological conditions on traffic flow. The experimental results show that the RMSE index of WSTAN-WF on the PeMSD4 and PeMSD8 datasets is reduced by an average of more than 7% compared with mainstream advanced models, verifying its superior prediction ability in complex traffic environments.