Abstract
Real-time traffic flow prediction can not only provide travelers withreliable traffic information so that it can save time, but also assist thetraffic management department to manage traffic system. It can greatly improvethe efficiency of the transportation system. Traditional traffic flowprediction approaches usually need a large amount of data but still give poorperformances. With the development of deep learning, researchers begin to payattention to artificial neural networks (ANNs) such as RNN and LSTM. However,these ANNs are very time-consuming. In our research, we improve the DeepResidual Network and build a dynamic model which previous researchers hardlyuse. Our result shows that our model has higher accuracy than somestate-of-the-art models. In addition, our dynamic model can perform better inpractical applications.