基于图卷积网络的路网短时交通流预测研究

智能交通系统是缓解交通拥堵行之有效的手段，精准的交通流预测是其实现的关键所在. 本文考虑路网拓扑结构和交通流时空相关性，提出基于图卷积网络(Graph Convolution Network，GCN)的大规模城市路网短时交通流预测模型，具有较高的预测精度、预测效率和现实解释意义；采用真实大规模城市路网浮动车数据对GCN模型进行测试，结果表明，GCN模型相对于现有模型，在预测性能上有较大提升.

Short-term Network-wide Traffic Prediction Based on Graph Convolutional Network

Intelligent transportation systems provide an effective means to alleviate traffic congestion. Traffic flow prediction is the key to realize it. This paper proposes a short-term traffic flow prediction model for largescale urban road networks based on graph convolutional network (GCN). The topological structure of the road network is considered as well as the spatial-temporal correlation of traffic flow, which results in high prediction accuracy, high efficiency, and interpretability of the model. A case study was performed on the model using realworld large-scale urban road network data. The results show that the GCN model greatly improves the prediction performance compared to existing benchmarks.