01100680

Component

Accident prediction models are usually developed using the negative binomial distribution, which results from a Bayesian Poisson-Gamma hierarchy to accommodate extra variation (over-dispersion). One of the most common and important predictors in such models is traffic volume which is known to be measured with uncertainty. Such errors-in-predictors can increase dispersion that may not be adequately treated using the negative binomial (instead of Poisson) regression. Improved estimates of traffic volume can be obtained from repeated observations of traffic flow which can be costly and not practical. In this paper, a less costly alternative is proposed which involves the use of a measurement errors model based on traffic flow time replicates. Such a model is then used in conjunction with the traditional negative binomial APM to circumvent the bias in predicting the aggregate number of accidents during the time period under study. The proposed approach was applied to a sample of accident, geometric and traffic volume data corresponding to rural 2-lane 210 road segments in British Columbia for the period of 2003-2005. The full Bayes method was utilized for parameter estimation, performance evaluation and inference through the use of Markov Chain Monte Carlo (MCMC) techniques. The results showed that the proposed approach provides an adequate fit to the data. It has also outperformed the traditional approach, which was found to significantly underestimate the predicted number of accidents in the presence of heavy traffic on long road segments. The paper concludes by identifying areas for further research.

01084478

08-1001

English

17p

2008

Transportation Research Board 87th Annual Meeting

DVD

Figures (1) ; References (30) ; Tables (3)

Bayes' theorem; Crashes; Markov chains; Mathematical prediction; Monte Carlo method; Simulation; Traffic flow

British Columbia

Data and Information Technology; Highways; Operations and Traffic Management; Planning and Forecasting; Safety and Human Factors; I72: Traffic and Transport Planning; I81: Accident Statistics

Transportation Research Board Annual Meeting 2008 Paper #08-1001

TRIS, TRB

Jan 29 2008 3:20PM