Route choice modeling with Support Vector Machine

Abstract

Abstract This paper aims at exploring the possibility of utilizing Support Vector Machine (SVM) to establish route choice model. A widely used non-parametric modelling approach, Neural Network, was used to compare with SVM. A stated preferences survey was conducted among 18 participants. Information about three route attributes including travel time, travel time fluctuations and fuel cost were given to participants for making route choice decisions. The data collected from the survey was used to calibrate and test both NN and SVM. The results show that SVM has similar prediction accuracy with NN but has much more computing efficiency.