Traffic-DesNet: A Robust Deep Learning-based Flow Prediction Model By The Using of DesNet

Real-time and accurate short-term traffic flow forecasting has become a critical problem in intelligent transportation systems (ITS). Short-term traffic flow is not only affected by time, but also by space. Ordinary short-term traffic prediction methods only solve the impact of time but not the impact of space. Convolutional neural networks can use convolution kernels to resolve spatial correlations and dependencies between time and space, so they can more accurately predict short-term traffic flows. We introduce DenseNet for short-term traffic flow prediction, use convolution kernels to extract features between time and space, and superimpose the features extracted from the upper layer to extract new features. It can effectively use high-level information to discover new features at the bottom again. To better capture the implicit correlation in some areas. It can be found through experiments that the model is not only superior in accuracy to traditional methods but also has a significant improvement in efficiency.