Short-term Traffic Flow Prediction Based on Deep Neural Network Considering Spatiotemporal Correlation
Download as PDF
DOI: 10.23977/iceccs.2018.036
Author(s)
Tian Yu, Weixiang Xu
Corresponding Author
Tian Yu
ABSTRACT
This study aimed to improve accurate of short-term traffic flow prediction. A considering spatiotemporal correlation traffic flow prediction method which is based on the deep neural network (ST-DNN) was proposed. In the aspect of temporal correlation, considering the influence of the lag operator and the seven days from Monday to Sunday on traffic flow prediction. In the aspect of spatial correlation, considering the impact on the current section traffic flow by the section of upstream and downstream of the road and the section with higher spatial correlation. Using data provided by Caltrans PEMS, results showed that the proposed model fits traffic flow prediction. In addition, a better performance compared with four existing methods, including ARIMA, shallow neural network (SNN), regression tree, and wavelet neural network. Clearly, this work has demonstrated the effectiveness of ST-DNN in the field of traffic flow prediction.
KEYWORDS
Short-term traffic flow prediction, Deep Neural Network, Spatiotemporal correlation