Bayesian Regression-Based Urban Traffic Models

Abstract

A Bayesian regression approach is used to develop equations for predicting travel time on central area streets with contributory variables that are intuitive and for which data are readily available in most transportation agencies. In development of multivariate regression models, two disparate sources of information are used: ( a) a priori (what is known before an experiment), and ( b) experimental data (information derived from an experiment). Output of traffic simulation obtained from NETSIM was used as the source of a priori information, whereas the experimental data were obtained from video recordings of traffic operations on selected central business district streets. Bayesian regression software was used in a systematic framework for predictive model development. The developed equations were assessed and results were interpreted from a Bayesian perspective in relation to the various model iterations attempted. The final models provide reasonable predictions of actual travel times that drivers would experience during peak traffic periods in medium to large central business districts.