Learning Traffic as Images: A Deep Convolutional Neural Network for Large-Scale Transportation Network Speed Prediction

摘要

This paper proposes a convolutional neural network (CNN)-based method thatlearns traffic as images and predicts large-scale, network-wide traffic speedwith a high accuracy. Spatiotemporal traffic dynamics are converted to imagesdescribing the time and space relations of traffic flow via a two-dimensionaltime-space matrix. A CNN is applied to the image following two consecutivesteps: abstract traffic feature extraction and network-wide traffic speedprediction. The effectiveness of the proposed method is evaluated by taking tworeal-world transportation networks, the second ring road and north-easttransportation network in Beijing, as examples, and comparing the method withfour prevailing algorithms, namely, ordinary least squares, k-nearestneighbors, artificial neural network, and random forest, and three deeplearning architectures, namely, stacked autoencoder, recurrent neural network,and long-short-term memory network. The results show that the proposed methodoutperforms other algorithms by an average accuracy improvement of 42.91%within an acceptable execution time. The CNN can train the model in areasonable time and, thus, is suitable for large-scale transportation networks.