Application of Bayesian model averaging for modeling time headway distribution

Time headway distribution plays a crucial role in traffic flow analysis, traffic state estimation, etc. Previous studies proposed different statistical distributions to model time headway. However, due to the model uncertainty, it is difficult to find certain types of distributions to describe time headway under different traffic conditions. To overcome the model uncertainty, a Bayesian model averaging (BMA) approach is applied to consider the advantages of different distributions to model time headway. Six time headway datasets were collected from two different traffic facilities (i.e., intersection and tunnel) in Guangzhou, China. A Markov Chain Monte Carlo (MCMC) sampler is adopted to estimate the unknown parameters and marginal likelihoods of candidate distributions in model space, which is determined by some commonly used time headway statistical distributions. The findings of the study illustrate that there is no single distribution (e.g., log-normal distribution, burr distribution, etc.) is universally appropriate for describing all time headway datasets, while BMA approach can accurately describe time headway distribution characteristics under different traffic conditions.