Comparative evaluation of Logit and Fuzzy Logic models of gap-acceptance behavior

In this study two alternative models of gap-acceptance behavior at unsignalized intersections with priority control are compared based on their ability to predict actual diver choices. The rst is a probabilistic model of the Logit type, while the second is a model based on Fuzzy Logic. The explanatory variables included in both models are the size of the time interval evaluated by drivers in their acceptance/rejection process, the total (queuing plus stop-line) delay experienced on the controlled approach, and the type of interval being evaluated (gap or lag). Both models are estimated/identied based on eld observations of actual gap-acceptance behaviour. The predictive power of the two models is compared using a technique known as ROC (Receiver Operating Characteristic) curve analysis, which apparently has never been applied before in the area of transport modeling. Our preliminary results are that the ROC curves suggest a slight superiority of the Logit model over the Fuzzy Logic model; however, other statistics, computed in order to provide a more complete comparative analysis of the two models, reveal no clear dominance of either model over the other.