Traffic Flow Prediction Based on Delayed Attention Neural Ordinary Differential Equation

To address the issues of delayed effects and insufficient exploration of spatiotemporal correlation characteristics in traffic flow prediction, a dynamic graph neural delay differential equation model is proposed to improve prediction accuracy. First, traffic flow features across different periods are extracted. Next, the delay time between upstream and downstream roads is calculated, and a delay differential equation is introduced to capture the time-lagged characteristics among nodes in the road network. Finally, validation on real-world public traffic flow datasets, including PEMS04 and PEMS08, demonstrates that the proposed model achieves an average reduction of approximately 35% in mean absolute error (MAE) and 27% in root mean square error (RMSE).