Traffic Flow Prediction with Compact Neural Networks

Abstract

As a crucial role of the Intelligent Transportation Systems, real-time traffic flow prediction technology is increasingly required for society. Traffic flow prediction is based on historical traffic data to predict upcoming traffic flow. In this paper, we employ Recurrent Residual Network and Temporal Convolutional Network to predict the traffic flow and compared it with several deep learning models such as Gated Recurrent Units and Long Short Term Memory in traffic flow prediction tasks. To the best of our knowledge, this is the first time that Recurrent Residual Network and Temporal Convolutional Network are applied to traffic flow prediction. We show that Temporal Convolutional Network has similar performance to other models with less computing parameters, which is simpler and clearer.