Abstract:In the Intelligent Traffic System (ITS), accurate prediction of the state of the traffic at the next moment is of great significance to transportation planning. Existing researches mainly focus on the research of the topological structure of the road network in space, and consider the temporal dependence at the same time. However, we have noticed that it is not only important to consider the dependence of time and space at the same time, but also other organizational relationships between the road network will also affect the forecast results. In this paper, we reconsider the correlation between roads and capture their correlation in both space and time. More specifically, we first encode the road network into two graphs (connect graph and similar graph) based on the connectivity and historical pattern similarity, and merge the graphs to make the connected edges carry more information. Then graph convolution is used for the fused graph. In order to capture the temporal dependence, we use one-dimensional convolution to first convolve the information before the graph convolution, and then perform the one-dimensional convolution on the information after the fusion graph, which can achieve fast spatial-state propagation from graph convolution through fast spatial-state propagation. We evaluate the predictive performance of our model by real-world traffic dataset and experiments prove that the addition of multi-graph is effective. Compared with the relatively new baseline, RMSE has dropped by about 10\(\%\).

Traffic flow prediction based on temporal multigraph convolutional neural network

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction

MGCN: Dynamic Spatio-Temporal Multi-Graph Convolutional Neural Network

Traffic Prediction Based on Multi-graph Spatio-Temporal Convolutional Network

MD-GCN: A Multi-Scale Temporal Dual Graph Convolution Network for Traffic Flow Prediction

A deep learning traffic flow prediction framework based on multi-channel graph convolution

Attention Mechanism Based on Temporal Graph Convolutional Neural Network for Traffic Flow Prediction

Predicting Traffic Flow Using Dynamic Spatial-Temporal Graph Convolution Networks

MFDGCN: Multi-Stage Spatio-Temporal Fusion Diffusion Graph Convolutional Network for Traffic Prediction

Multi-graph Spatio-temporal Graph Convolutional Network for Traffic Flow Prediction

MTGCN: Multi-graph Fusion Based Temporal-Spatial Convolution for Traffic Flow Forecasting

Temporal Graph Convolutional Network for Urban Traffic Flow Prediction Method.

STGMN: A gated multi-graph convolutional network framework for traffic flow prediction

Traffic Flow Prediction Based on Deep Neural Networks

Dynamic multi-graph convolution recurrent neural network for traffic speed prediction

Cross-Attention Fusion Based Spatial-Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction

Traffic Flow Prediction via Spatial Temporal Graph Neural Network

Spatial–Temporal Traffic Prediction Model Based on Adaptive Graphs Fusion and Dual-Graph Collaborative Convolution

Geographic-Semantic-Temporal Hypergraph Convolutional Network for Traffic Flow Prediction.

Traffic Flow Prediction Via Spatial-Temporal Geometric Graph Convolutional Network