In general terms, discrete choice models are calibrated using data obtained from Revealed Preference (RP) and Stated Preference (SP) surveys. In transportation planning, one of the main sources of data is the Origin/Destination (O/D) Survey, which is an R

Abstract

In general terms, discrete choice models are calibrated using data obtained from Revealed Preference (RP) and Stated Preference (SP) surveys. In transportation planning, one of the main sources of data is the Origin/Destination (O/D) Survey, which is an RP survey and describes the actual choices and behaviors of individuals. However, it is not possible, through this source, to characterize the alternatives not chosen. This study has two related aims: (1) to propose a criterion to characterize the travel mode alternatives using RP data, and (2) to test the improvement of travel mode choice estimates based on including characteristics of alternatives. First, the CART (Classification and Regression Tree) algorithm was used to characterize the travel times of the travel modes available in the study area (city of São Paulo, Brazil). The trips were classified according to independent variables selected by the algorithm, and average travel time values were obtained for five travel mode alternatives – information not previously available in the RP survey. Finally, the improvement of discrete choice modeling, based on including average travel times, was tested using a validation sample and performance metrics, such as Hit rates and LogLikelihood values. An increase in estimates was observed from including travel duration, and the proposed method is an academic contribution to the modeling based on RP data.