A method to directly derive taste heterogeneity of travellers’ route choice in public transport from observed routes

The heterogeneity of passengers’ route choice has been explained by randomizing the parameters, also known as taste parameters, that determine the way the attributes are relatively weighed in the disutility he/she perceives from a route. Growing availability of massive route choice data from, e.g. GPS or Smart Card system has made expected a model that derives the distribution of taste parameters from RP-data rather than relies on a prescribed distribution. This study availed itself of the intensive set of route choice data from Smart Card system as well as inverse optimization to calibrate the joint pdf of taste parameters to best signify the user-optimality of observed routes. Tested on 5 daily sets of real route choice, which amounts to 50,000 trips from the metro of Seoul metropolitan area, the proposed model notably enhanced the predictability compared to the previous models adopting a mixed-logit-based SUE or a non-parametric estimation method.