Traffic Prediction Based on Adaptive Spatiotemporal Graph Neural Network
Accurate traffic prediction is of great significance to urban planning and traffic safety.Most of the existing predictive models focus on designing complex graph neural network structures,with the help of predefined graphs to capture the characteristics of traffic data.However,traffic data has a strong spatial dependency,which means that there are often complex correlations between nodes of a road network topology map,and the topology map of a road network changes over time.To solve this problem,an adaptive spatiotemporal graph neural network is proposed,and a graph structure learning component is first put forward,which captures the macro and micro information of the transportation network respectively and integrates them into an optimal graph adjacency matrix.Then,a spatiotemporal convolution block is designed to capture the spatiotemporal characteristics of traffic data.Experiments are carried out on the METR-LA and PEMS-BAY datasets,and the experimental results show that the predictive performance of the proposed model is better than that of the mainstream model.
deep learningtraffic predictiongraph neural networkspatiotemporal convolution block
NETL
NSTL
万方数据
