Models to Predict Traffic Volatility in Transportation Networks

This paper describes the application and relative performance of three different models for predicting traffic volatility in transportation networks. The Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model, the Stochastic Volatility (SV) model and the Realized Volatility (RV) model are implemented in a real urban arterial network using real-time traffic data of volumes and occupancies. The experimental results provide evidence of the superior performance of the SV model and, at a lesser extent, of the RV model to produce out-of-sample volatility forecasts, in comparison to the standard GARCH methodology.