Traffic parameters prediction using a three-channel convolutional neural network

Abstract

Traffic three elements consisting of flow, speed and occupancy are very important parameters representing the traffic information. Prediction of them is a fundamental problem of Intelligent Transportation Systems (ITS). Convolutional Neural Network (CNN) has been proved to be an effective deep learning method for extracting hierarchical features from data with local correlations such as image, video. In this paper, in consideration of the spatiotemporal correlations of traffic data, we propose a CNN-based method to forecast flow, speed and occupancy simultaneously by converting raw flow, speed and occupancy (FSO) data to FSO color images. We evaluate the performance of this method and compare it with other prevailing methods for traffic prediction. Experimental results show that our method has superior performance.