Discrete Choice Modeling for Modal Disaggregation of Mobile OD Flows: Comparison with Smart Card Data

Abstract

Estimating travellers’ preferences for different transport modes can be achieved using Discrete Choice Models (DCMs), which often rely on outdated travel surveys (HTSs) that misrepresent current urban mobility. Digital data can address this problem by complementing the survey, but challenges arise when HTS and digital sources are from distant periods and due to the incompleteness and bias inherent in digital data. In this paper, we propose a first step in a broader effort to update mode choice probabilities using recent digital traces. We train a DCM on the HTS to estimate mode choice probabilities, which are then used to disaggregate mobile phone OD matrices by mode. We investigate the extent to which preprocessed mobile phone data disaggregated with a DCM model is comparable to smart card data used to estimate OD matrices for the public transport mode. Experiments are conducted using real data collected in the Métropole de Lyon, France.