LAUSR

ABSTRACT Smart card data provides a new perspective for estimating a metro passenger’s path choice model in a large-scale urban rail transit network with multiple alternative paths between origin-destination pairs. However, existing research does not consider correlations of path travel times among alternative paths when using smart card data for estimation purposes, leading to biased estimations. This paper proposes an approach to estimating the path choice model considering path travel time correlations. In particular, a simplified form of measuring path travel time correlations caused by shared links is proposed to improve estimation efficiency. Then a framework for a linking path choice model and smart card data is developed based on a Gaussian mixture model; an expectation maximization-based estimation algorithm is also provided. Finally, taking the Guangzhou Metro in China as an example, the superiority of estimations based on smart card data considering correlations is observed in both statistical terms and predictions.