Urban Traffic Congestion Prediction and Dynamic Path Optimization Algorithm Based on Spatiotemporal Graph Neural Network

This paper proposes an urban traffic congestion prediction and dynamic path optimization algorithm based on spatiotemporal graph neural network, which combines the advantages of Spatiotemporal graph Convolution network (STGCN), Graph Attention Network (GAT) and Dijkstra algorithm. A new Graph Spatial-Temporal Dijkstra (GSTD) model is constructed. First, STGCN is used to capture the spatiotemporal dependence of the traffic network and predict the future traffic state. Secondly, the dynamic modeling of node relationships in the road network is carried out by GAT to enhance the understanding ability of complex road network structure. Finally, combined with Dijkstra algorithm, path planning is dynamically optimized based on prediction results. In the experiment, compared with the traditional route optimization methods, the experimental results show that the SGD model can efficiently and accurately predict traffic congestion and optimize route planning in real time, providing reliable technical support for urban traffic management. This fully verifies the reliability and effectiveness of GSTD architecture in urban traffic congestion prediction and dynamic path optimization algorithm.