Forecasting Freeway Traffic Volumes with Adverse Weather

Forecasting traffic volumes under adverse weather in advance contributes to allocating traffic resources for traffic managers and formulating optimal travel strategies for travelers, which assists in preventing and offsetting the impact of adverse weather on traffic. Consequently, the accurate prediction of traffic volume is vital. This study proposes an adverse weather traffic volume prediction model combining Convolution Neural Networks, Bi-directional Long Short-Term Memory, and Attention Mechanism. The 5-minute highway traffic volume data from December 1, 2021, to March 13, 2022, in Minnesota, U.S.A, and the weather data in the same period provided by Meso West were used as the experimental data. The proposed model was compared with three single prediction models, two validated hybrid models, and the model itself without integrating weather factors. The experiments show that the prediction accuracy of the proposed model is higher than other comparison models.