DLW-Net model for traffic flow prediction under adverse weather

Abstract

To predict traffic flow under adverse weather, a hybrid deep learning model concerning adverse weather (DLW-Net) is formulated. The DLW-Net model consists of the target and global analysis parts. For the target analysis part, the spatio-temporal characteristics of traffic flow data are analyzed using the convolutional neural network (CNN), the long short-term memory (LSTM) and gated recurrent unit (GRU) neural networks. For the global analysis part, the variation rules of traffic flow and weather data are extracted using the LSTM model. Additionally, the characteristics of traffic flow under normal and adverse weather are also discussed. The developed model is verified using three cases. The results show that traffic volume and speed would reduce under heavy rain compared to normal weather, however, drizzle has little impact on traffic flow patterns; the rules of traffic speed data are disturbed by strong wind; and the DLW-Net model performs best under all the conditions.

KEYWORDS:

• Traffic flow prediction
• adverse weather
• deep learning
• hybrid method
• CNN
• GRU

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data used to support the findings of this study are available from http://www.openits.cn/ and http://pems.dot.ca.gov/. The models or codes that support the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Funding

The research reported in this paper has been funded by the Open Foundation of Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport of China Academy of Transportation Sciences [2020B1203], and partly sponsored by the National Natural Science Foundation of China [52172314] and the Fundamental Research Funds for the Central Universities of Ministry of Education of China [DUT20JC40].