The aim of this study is to produce sensible forecasts for a number of combinations of classes such as age/gender, transport mode, road type, rush hour or not, and accident severity. The Mobility Explorer imposes a number of conditions which the traffic safety component has to satisfy. One of these conditions is the development of forecasts for combinations of classes of road users or traffic. The study started by investigating the disaggregation level made possible by the data. Attention has been paid to the reliability of the results. This has been done: (1) by comparing the observed numbers with the forecasted figures; and (2) by computing confidence intervals (a measure of reliability) for the forecasts, assuming the correctness of the applied model. Finally, it is described how to improve the measure of reliability: (i) by incorporating knowledge about the uncertainty in the mobility information; and (ii) by estimating the deviance of the used model from the 'true' model.
Abstract