Prediction of Traffic Conflict in Freeway Merging Area Based on Bayesian Model

Abstract

Aiming at the problems of complicated traffic flow and high accident safety risks in the expressway merging area, considering the discrete and heterogeneous characteristics of traffic conflict data, a Poisson-lognormal distribution model (PLN) and the random parameters Poisson-lognormal traffic conflict model (RP-PLN) were developed; The posterior distributions of the models parameters were estimated by Bayesian method and the Markov chain Monte Carlo (MCMC) simulation. The goodness-of-fit of models were compared by using the deviance information criterion. The results show that the goodness of fit of the random parameters Poisson-lognormal traffic conflict model (RP-PLN) is higher than that of the Poisson-lognormal distribution t model (PLN).