In recent years, the development of big data acquisition and storage, computer technology and communication technology has provided new momentum for ITS, while traffic speed prediction is a core link of ITS. In order to achieve large-scale traffic forecasting in urban road network and extract the time series feature and spatial feature of road network speed evolution, a spatiotemporal prediction method based on 3D convolution neural network is proposed in this paper, using gridded historical traffic data and corresponding road network traffic speed for training. Finally, in the empirical analysis stage, 3D CNN is evaluated and compared with the prediction results of 2D CNN, LSTM, and BPNN models on the whole, midweek and weekend. Experimental results show that the MAE, MAPE, and RMSE indices of the test set are at least 10% better than other models. It has a good performance in the actual road network traffic speed prediction.
Large-Scale Spatiotemporal Prediction Method of Traffic Speed Based on 3D Convolutional Neural Network
20th COTA International Conference of Transportation Professionals ; 2020 ; Xi’an, China (Conference Cancelled)
CICTP 2020 ; 163-172
09.12.2020
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Spatiotemporal Patterns in Large-Scale Traffic Speed Prediction
Online Contents | 2014
|DSpace@MIT | 2014
|Multi-scale traffic flow prediction method based on graph convolutional neural network
Europäisches Patentamt | 2021
|