Hierarchical Fuzzy Inference System to Evaluate Expert Opinions on Median Safety

Abstract

Transportation experts may be asked to provide an opinion about a topic for various reasons. Most studies use simple statistical methods to aggregate these opinions. However, because experts use linguistic information and their own subjective decision criteria to formulate and express their opinions, this aggregation can ignore the unique characteristics of individual experts and the decision criteria they used. To overcome these limitations, a hierarchical fuzzy inference system was developed to evaluate the subjective opinions from transportation experts on median safety. Variables were selected from the results of a survey of transportation experts conducted as part of a previous study. The fuzzy membership functions for six selected variables were constructed through a review of the experts’ opinions and transportation safety references. The proposed fuzzy inference system was decomposed into two systems with a hierarchical structure to reduce the complexity of the multivariable fuzzy system. In the decomposed system, the degree of median safety, distilled from geometric and traffic flow conditions, was represented by a fuzzy median safety index (FMSI). Through comparison with observed median crashes, it was found that the mean cross-median collision (CMC) frequency and mean CMC crash rate increased exponentially with an increase of FMSI, typically for roadway segments with an FMSI equal to or greater than 0.7. The fuzzy inference system can then be used to estimate the likelihood of a median crash occurring within an individual segment of highway.