A Bayesian Network Model for the Probabilistic Safety Assessment of Roads

Abstract

A Bayesian network model for probabilistic safety analysis of roads and highways is introduced. After indicating how the list of variables and the conditional probability tables of the Bayesian network model are built, based on a video of the road, a short discussion about how maximum likelihood and Bayesian network methods can be applied to estimate the model parameters using standard methods. Next, a partitioning technique is suggested to convert the non-linear problem of computing marginal and conditional probabilities after evidence into a problem whose complexity becomes linear in the number of variables. Finally an example of application is used to illustrate the proposed methodology and some conclusions are drawn.