Short Term Traffic Flow Prediction Based on Deep Learning

Abstract

In this paper, three traffic prediction models based on deep learning are used to predict the traffic flow of capital airport. First, we reconstruct the washed traffic flow data to make the prediction results spatial-temporal. After smoothing and standardization, the characteristics of airport traffic data are studied using the stacked automatic coding machine (SAE) model, the long and short memory network (LSTM) model and the control gate recursion (GRU) model, and the final results are predicted by using the regression layer on the top layer. Finally, the results are obtained by anti-standardization, and the three models are obtained. We then compared the reliability of the three models and proved different loss functions.