<p>Traffic flow prediction is a vital part of an intelligent traffic management system. The critical challenge of the traffic flow prediction task is to fully use observable historical information to extract the hidden spatiotemporal dependence. For this reason, a synchronous spatiotemporal grammar graph attention network is proposed based on multi-dimensional edge information (MDEI-SSTGGAT) to achieve short-term traffic flow prediction tasks. In the designed model, the spatial nodes are first clustered. Then, a local spatiotemporal graph is established for each category for extracting synchronous spatiotemporal features. In the feature extraction module, the graph attention network with multi-dimensional edge information captures the spatiotemporal features dynamically, and the grammar graph structure fuses the hidden features of three observable traffic parameters. The model fully uses observable traffic parameter features to improve the prediction accuracy while keeping the computation time short. Simulation results on real data sets show that the prediction performance of this model is better than the existing prediction methods.</p>