Comparing Direct Transferability of Logit and Fuzzy Logic Models of Gap Acceptance at Unsignalized Intersections

This paper presents a comparative analysis of the performance of Logit models vs. Fuzzy logic models of gap-acceptance behavior in terms of their direct transferability. This refers to situations in which a model estimated in a given experimental context is directly applied in a different context without any updating. Data collected at four priority intersections having different characteristics in terms of geometry, location and type of control (stop vs. yield sign) were used to test the quality of the results obtained when transferring each type of model from an original to an alternative situation. The comparison was carried out based on metrics commonly used in the so-called ROC (Receiver Operating Characteristic) curve analysis. The results show that both models have essentially the same capability in terms of direct transferability, and that their performance in contexts different from those of original development could be more than adequate for application purposes.