Development and Evaluation of Recurrent Neural Network based Models for Hourly Traffic Volume and AADT Prediction

The prediction of high-resolution hourly traffic volumes of a given roadway is essential for future planning in a transportation system. Traditionally, Automatic Traffic Recorders (ATR) are used to collect this hourly volume data, which is often used to predict future traffic volumes accurately. However, these large data sets generated from these various ATRs are time series data characterized by long-term temporal dependencies and missing data sets. Regarding the temporal dependencies, all roadways are characterized by seasonal variations that can be weekly, monthly or yearly, depending on the cause of the variation. Regarding the missing data in a time-series sequence, traditional time series forecasting models perform poorly under their influence. Therefore, a robust, Recurrent Neural Network (RNN) based, multi-step ahead time-series forecasting model is developed in this study. The simple RNN, the Gated Recurrent Unit (GRU) and the Long Short-Term Memory (LSTM) units are used to develop the model and evaluate its performance. Two novel approaches are used to address the missing value issue: masking and imputation, in conjunction with the RNN unit. Six different imputation algorithms are then used to identify the best model. Our analyses indicate that the LSTM model performs better than simple RNN and GRU models, and imputation performs better than masking. The performance of imputation methods is also dependent on the percentage of missing data in the input dataset. Overall, the LSTM-Median model is deemed the best model in all scenarios for AADT prediction, with an average accuracy of 98.5%.