基于GCN-AGRU的短时交通流预测OA

In view of the problem that the existing short-term traffic flow prediction methods do not consider the frequency-domain characteristics of traffic flow,a new GCN-AGRU(Graph Convolutional Neural Network-At-tention Gated Recurrent Unit)short-term traffic flow prediction model is constructedr.Combining the GCN(Graph Convolutional Neural Network)network based on adaptive graph structure with the GRU(Gated Recurrent Unit),and conducting frequency-domain analysis of the hidden state of the GRU by using FCA(Frequency Channel Atten-tion)attention,the periodic characteristics in traffic flow can be captured more accurately.Experiments were con-ducted based on the Wuhan Road network dataset and the PeMS04 dataset,and the prediction results of other mod-els were compared and analyzed.Compared with the optimal baseline model,the MAE(Mean Absolute Error)、RMSE(Root Mean Square Error)and MAPE(Mean Absolute Percentage Error)of the proposed model decreased by 9.84 percentage points,1.87 percentage points and 4.88 percentage points respectively,significantly improving the accuracy and stability of short-term traffic flow prediction.