基于GCN-Transformer模型的短时交通量确定性预测

Deterministic Prediction of Short-Term Traffic Volume Based on GCN-Transformer Model

摘要: 为提高短时交通量的预测精度，充分利用交通网络的拓扑结构与时序信息，本文提出了一种基于GCN-Transformer模型的短时交通量确定性预测。首先读取并预处理四个方向的交通流量数据，构建路网结构邻接矩阵，划分数据集。然后利用GCN提取特征并进行图卷积操作，将输出作为Transformer模型的输入来学习时序信息。最后使用训练好的模型进行预测，并计算评价指标。研究结果表明，本文所提出的混合预测模型GCN-Transformer拥有较好的结果，在采用的四个评价指标中，RMSE、MAE、RMSPE和MAPE的值分别为9.886%、7.977%、11.018%、11.734%。

Abstract: In order to improve the prediction accuracy of short-term traffic volume and make full use of the topology and timing information of traffic network, a deterministic prediction of short-term traffic volume based on GCN-Transformer model is proposed in this paper. Firstly, the traffic flow data in four directions are read and preprocessed, the network structure adjacency matrix is constructed, and the data set is divided. Then, GCN is used to extract features and perform graph convolution operations, and the output is used as the input of Transformer model to learn timing information. Finally, the trained model is used to predict and the evaluation index is calculated. The research results show that the hybrid prediction model GCN-Transformer proposed in this paper has better results, and the values of RMSE, MAE, RMSPE and MAPE among the four evaluation indicators are 9.886%, 7.977%, 11.018% and 11.734%, respectively.