Abstract
This paper presents a methodology to estimate confidence intervals of forecast corridor travel times. First the mean link travel time forecasts are derived using spectral basis neural networks. A Bootstrap is then used to estimate the variance associated with these forecasts. A Taylor series approximation approach is adopted to approximate the expected mean and variance of the corridor travel time forecasts from the forecast of the mean and variance of the link travel times. Based on these estimates, confidence intervals are calculated. Preliminary results using the automatic vehicle identification (AVI)-based travel time from Houston, Texas are presented. (A*)