Avoiding the crowds: understanding Tube station congestion patterns from trip data

Abstract

For people travelling using public transport, overcrowding is one of the major causes of discomfort. However, most Advanced Traveller Information Systems (ATIS) do not take crowdedness into account, suggesting routes either based on number of interchanges or overall travel time, regardless of how comfortable (in terms of crowdedness) the trip might be. Identifying times when public transport is overcrowded could help travellers change their travel patterns, by either travelling slightly earlier or later, or by travelling from/to a different but geographically close station. In this paper, we illustrate how historical automated fare collection systems data can be mined in order to reveal station crowding patterns. In particular, we study one such dataset of travel history on the London underground (known colloquially as the "Tube"). Our spatio-temporal analysis demonstrates that crowdedness is a highly regular phenomenon during the working week, with large spikes occurring in short time intervals. We then illustrate how crowding levels can be accurately predicted, even with simple techniques based on historic averages. These results demonstrate that information regarding crowding levels can be incorporated within ATIS, so as to provide travellers with more personalised travel plans.