Mode Choice Behavior Modeling: A Synergy by Hybrid Neural Network and Fuzzy Logic System

Mode choice modeling is one of the crucial measurements in transportation policies as it helps in formulating city transport policies. Travel to school mode choice modeling was a focus of many researchers to study the effect of different factors on students' mode choice decisions. On this backdrop, this study investigated the effect of fuzzy c-means (FCM) and subtractive clustering (SC) algorithms on the performance of the adaptive neuro-fuzzy inference system (ANFIS) in predicting mode choice behavior of school’s students. This modeling approach generates a synergy between neural network and fuzzy logic models by representing the fuzzy inference system into the framework of a neural network and adopting a clustering algorithm to select the number of rules of the fuzzy logic system. This approach addresses the shortcomings of the fuzzy logic system pertinent to the number of rules. Two models, namely FCM-ANFIS and SC-ANFIS, were developed using school level, travel distance to school, monthly income, travel time, family size, number of students per family, and parents’ educational level as inputs. Classification accuracy and F1 score were used as measures to make a comparison between both models. The results revealed that SC-ANFIS and FCM-ANFIS performed well in predicting mode choice to school. The results of the F1 score and prediction accuracy showed that the SC-ANFIS outperformed the FCM-ANFIS in predicting the walking and passenger car modes of school students. The modeling framework used in this research presents a new methodology that is not extensively used to solve this kind of problem.