Supporting data for the manuscript "Dynamic Graph-Attention Traffic Network for Enhanced Urban Traffic Flow Prediction".

Taking center stage in current traffic flow forecasting research are the PEMS04 and PEMS08 datasets (https://planet.osm.org), which are derived from the California Transportation Management System's (Caltrans) extensive collection efforts of freeway induction coil detectors. These datasets not only provide traffic speed data updated every 5 minutes, but also meticulously record multi-dimensional information such as traffic flow and vehicle occupancy, reflecting real-time traffic conditions on some of California's key freeways. This rich collection of data, which covers a diverse range of traffic flow patterns across seasons, weekdays, and holidays, provides researchers with valuable experimental materials to explore the complex dynamics of traffic flow.