Spatio-Temporal Investigation of Public Transport Demand Using Smart Card Data

Policymakers must find efficient public transport solutions to promote sustainability and provide efficient urban mobility in the course of urban growth. A growing number of research papers are applying Geographically weighted regression (GWR) to model the relationship between public transport demand and its influential factors. However, few studies have considered the rapid development of journey inference from ticket transaction data. Similarly, the potential of GWR to analyze spatio-temporal changes that reflect changes in transportation supply and thus provide a measure for evaluating the local success of transport supply changes has yet to be exploited. In this paper, we use inferred journeys from smart card inferences as the dependent variable and analyze how public transport demand responds to a set of explanatory variables, emphasizing transport supply. Consequently, GWR and its successor Multiscale Geographically Weighted Regression (MGWR) are applied to analyze the spatially varying impact of transport supply changes for seven consecutive time frames between autumn 2017 and spring 2020, allowing conclusions about local changes in transport demand, as well as the benchmarking of transport supply changes. The (M)GWR framework’s predictive power is evaluated by training the model with past transport supply data and testing the model with data from the following consecutive years. The conducted analyses reveal that the (M)GWR model, using inferred journeys and transport supply data, can retrospectively predict the impact of transport supply changes on travel behavior and thus provides conclusions about the success of transport policies.