Traffic Flow Prediction with Conv-SAE

Abstract

As traffic jams become more serious, accurate traffic flow forecasting is essential to ease traffic pressure. In order to meet the needs of traffic forecasting, this paper proposes a combination model Conv-SAE based on convolution and SAE(stacked autoencoders), which roughly extracts the spatial features and temporal features by the SAE module, and then fully extracts the spatial features through the convolution module. The experimental results show that the prediction accuracy of the method we use is more competitive than other models.