User Cost Estimation on Road Networks by Means of Bayesian Probabilistic Networks

In this paper a methodology for the development of multiple impact models for road networks is described. Road impacts can be categorized into impacts for different stakeholder groups, namely the public, the owner and the road users. The current investigations address multiple impacts only for road users, but the developed methodology can be extended to estimate the impacts of all stakeholder groups. Impacts for road users are differentiated into costs due to travelling time, vehicle operation and accident injuries. The accident and injury costs are assessed based on multivariate regression analysis and Bayesian Probabilistic Networks. The proposed methodology for the assessment of road user impacts is different from existing impact models since uncertainties are incorporated into the model. Accordingly, all variables of the model are represented probabilistically. To demonstrate the usefulness of the methodology, models were developed for the prediction of multiple road user impacts on three road segments that were different in terms of traffic configurations, road designs and surface conditions. Based on the assumptions made for the model development, the results show that accident and injury costs represent only a small share of the total user impacts. An important feature of the introduced methodology for road impact assessment is that it provides decision support, e.g. on how to optimally allocate budgets into accident risk reducing interventions and evaluate the portfolio of changeable measures in terms of their effect on the accident risk before and after they are implemented.