Macroscopic Traffic Stream Variable Prediction with Weather Impact using Recurrent Learning Approach

Accurate prediction of the macroscopic traffic stream variables such as speed and flow is important for traffic operation and management in an intelligent transportation system. The accurate prediction of these variables is challenging because of the non-linear and complex characteristics of the traffic stream. With recent computational technologies and huge data availability, such a problem is solved using data-driven approaches. Traditional data-driven approaches use shallow architecture which ignores the hidden influencing factor and proved to have limitations in a high dimensional traffic state. Adverse weather conditions like fog, wind, rainfall, and snowfall affect the visibility of the driver, mobility of vehicle and road capacity. We examine the effect of rainfall on traffic stream variable prediction. The deep learning approaches use a layered architecture to extract the inherent features in the data. In this paper, recurrent neural network(RNN) and its variant long short term memory(LSTM) are used and their accuracy of predicting the traffic stream variables with and without rainfall variable is studied. To validate model performance, traffic sensor data from an arterial road and rainfall data from weather stations in San Diego are used for model training and testing. The test experiments show that with the combination of traffic data and rainfall data, recurrent learning models give better prediction accuracy over the model without rainfall data. Also, the LSTM outperforms other deep learning models in the presence of rainfall data.