Full Bayesian Method for the Development of Speed Models: Applications of GPS Probe Data

Abstract

Traffic speed is one of the basic variables that indicates the level of service of a road entity. It plays an essential role in transportation planning and management. This study attempts to establish a prediction model for speed distribution, in terms of average travel speed and standard deviation, using probe vehicle data in Hong Kong. Taking advantage of detailed traffic flow data obtained from the annual traffic census, a comprehensive traffic information database can be established using the geographical information system technique. The effects of traffic flow, road geometry, and weather conditions on speed distribution are determined using the Markov-chain Monte Carlo (MCMC) simulation approach full Bayesian method.