Estimating Public Transport Demand Information Using Crowdsourced Data

Abstract

The analysis of transit demand and its complex dynamics has typically relied on survey-based data that captures only a small fraction of the total demand. Recently, emerging data-driven approaches have been applied to transportation issues and these typically rely on sensing data gathered by mobile devices under the so-called mobile crowdsensing (MCS) paradigm. This type of data can be a powerful source of information especially in areas where transit data is not available. This work aims to investigate the possibility of using Google Popular Times (GPT), a widely available crowdsensed data, to estimate the passenger flows of individual subway stations. Our results show that we can estimate precisely both entrances and exits profiles, which is particularly challenging because GPT only provide popularity trends. Our analysis is carried out on more than 105 subway stations of Manhattan and it is validated using turnstile count data from the stations.