A big data approach to understanding pedestrian route choice preferences: Evidence from San Francisco

Big data from smartphone applications are enabling travel behavior studies at an unprecedented scale. In this paper, we examine pedestrian route choice preferences in San Francisco, California using a large, anonymized dataset of walking trajectories collected from an activity-based smartphone application. We study the impact of various street attributes known to affect pedestrian route choice from prior literature. Unlike most studies, where data has been constrained to a particular destination type (e.g. walking to transit stations) or limited in volume, a large number of actual trajectories presented here include a wide diversity of destinations and geographies, allowing us to describing typical pedestrians' preferences in San Francisco as a whole. Other innovations presented in the paper include using a novel technique for generating alternative paths for route choice estimation and gathering previously hard-to-get route attribute information by computationally processing a large set of Google Street View images. We also demonstrate how the estimated coefficients can be operationalized for policy and planning to describe pedestrian accessibility to BART stations in San Francisco using 'perceived distance' as opposed to traversed distance.