Abstract

Traffic flow prediction effectively supports the sustainable expansion and operation of modern transport networks, one of the emerging research areas in intelligent transportation systems. Currently, most common traffic flow prediction methods use deep learning spatial–temporal models based on graph convolution theory, which cannot deeply explore the spatial hierarchy and directional information of traffic flow data due to their structural characteristics. To address this problem, a spatial–temporal neural network based on tree structure (TS-STNN) is created to anticipate future traffic flow at a specific time at a target location. The principle of this method is to use the characteristics of the tree structure to construct a plane tree matrix with hierarchical and directional features, which is finally fused into a spatial tree matrix to extract the spatial information. Meanwhile, the temporal correlation of traffic flow data in the traffic network is analyzed by TS-STNN using Gated Recurrent Units (GRUs). By comparing with the existing baseline methods, it is verified that the TS-STNN model has high prediction accuracy in both Random Uniformly Distributed (RND) and Small-Scale Aggregation of Node Distributed (SSAND) scenarios. It is further demonstrated through ablation experiments that the developed tree convolution module greatly impacts the TS-STNN accuracy.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call