Analysis and Prediction of Ridership Impacts during Planned Public Transport Disruptions

TL;DR

This paper analyzes how planned public transport disruptions affect passenger demand, estimates demand elasticities for different groups, and develops a neural network model to accurately predict ridership impacts during closures, aiding better resource planning.

Contribution

It introduces an empirical analysis of demand response to planned disruptions and presents a neural network model for precise ridership prediction during closures.

Findings

Demand response is lower for frequent users and during peak weekdays.

Weekend leisure journeys show higher demand elasticity.

The neural network model predicts demand with high accuracy.

Abstract

Urban metro and tram networks are regularly subject to planned disruptions, including closures, resulting from the need to maintain and renew infrastructure. In this study, we first empirically analyse the passenger demand response to planned public transport disruptions based on individual passenger travel behaviour, based on which we infer generalised journey time and cost elasticities for different passenger groups and time periods of the day. Second, we develop a model which enables predicting public transport demand for individual origin-destination pairs affected by a closure. The model is trained based on the empirically observed travel behaviour. The proposed method is applied to a case study closure in Amsterdam, the Netherlands, based on which we empirically derive generalised journey time and generalised journey cost elasticities. Our results suggest that passengers demand response is lower for frequent users of the public transport network, as well as during weekdays, especially during the peak periods. Arguably, this stems from a higher share of captive passengers with a mandatory journey purpose in these segments, who will continue making their journey nevertheless. During weekends, with typically higher shares of leisure related journeys, a much more pronounced demand response is found. The estimated neural network regression model is able to predict passenger demand during public transport closures with a high level of accuracy. This provides public transport agencies more precise insights into the impact of closures on their revenue losses and on the potential need for resources reallocation.