Weather-Aware Data-Driven Traffic Speed Prediction for Suburban Road using deep learning models

Traffic speed prediction, as a vital component of intelligent transportation management systems, can significantly enhance traffic efficiency and safety.The aim of this study is to examine meteorological information in order to preserve the dynamic nature of traffic speed and to improve the predictive capabilities of existing models for traffic speed forecasting, with the Chalus Road considered as the case study.Three hybrid LSTM-based models, including BI-LSTM and CONV-LSTM, were employed to predict traffic speed using weather-related information. These models were accurately compared based on performance metrics. The accuracy of the predictive models was evaluated by comparing the actual traffic speed values with the predicted values and their associated prediction errors. The results indicate that the CONV-LSTM model outperforms the others, achieving a traffic flow error of 0.75 vehicles. Furthermore, the findings reveal that the wet-bulb temperature variable has the most significant impact on traffic speed along the Chalus Road.The study also demonstrates that, depending on the model’s data type, these models can be utilized either independently or in combination with other models to achieve better performance.This innovation holds substantial potential for enhancing management and planning of intercity highways, thereby improving traveler comfort and road safety.