Urban traffic flow forecast based on dual path network

Abstract

Traffic flow prediction is very important for city construction.Time and space factors have a great impact on traffic flow, Traditional traffic prediction methods can only capture temporal correlations and not capture spatial and regional correlations. The use of convolutional neural networks can well capture the correlation between regions and the dependence between time and space, which can make traffic prediction more accurate. Therefore we introduced the dual-path network, and divided the traffic profile after convolution into two paths and trained at the same time. One path is ResNet and one path is DenseNet. The dual path network combines the advantages of these two networks for traffic prediction. The experimental results show that compared with the traditional traffic prediction model, the model not only improves the efficiency of the network but also improves the prediction accuracy of the network.