Additive Bayesian Networks for Driving Risk Analysis in Road Safety

Additive Bayesian networks (ABNs) are a form of graphical models that extend the conventional generalized linear model (GLM) to multiple depend-ent variables through the representation of joint probability distribution. Due to their versatility, ABNs have been extensively utilized in various studies. How-ever, their application in road safety is still in its early stages, with initial studies focusing on the analysis of crashes involving automated vehicles for instance. To the best of our knowledge, our study is the first to apply ABNs to a case study on driving risk. In this paper, we present a case study where ABNs are applied to examine multivariate data related to road safety with a risk analysis. The ABN model not only identifies statistically significant associations but also distin-guishes between variables that are directly or indirectly dependent on one or more outcome variables.