This thesis contributes to the research area of route choice estimation with smart card data in large metro networks by addressing the issues with finite mixture models.
The motivation for this research comes from the problem that public transport authorities need to know passengers’ route choice for their key functions. Recently, many cities adopted smart cards, which produced a wealth of data for researchers. However they reveal only the entry/exit station, not the chosen route.
Within the scope of this research is to address the following research problems:
Firstly, to propose a model that generates automatically the route choice set for all types of OD pairs in a metro network by finding a set of shortest routes with the K shortest path algorithm, and narrowing down this set by applying the generalised cost proportion of routes as the attribute cut-off.
Secondly, to introduce the concept of superstations by grouping those stations from/to which passengers have similar route choice patterns; and to aggregate the Observed Journey Times (OJT) of station-to-station OD pairs, so that the finite mixture model can be applied on a larger dataset.
Thirdly, to investigate the question of fail-to-board delays in two aspects: considering that at different origin stations, the fail-to-board delays may be different; as well as updating the route choice estimates, with the information on the fail-to-board delays along different routes.
The methodologies are illustrated through the case studies on the London Underground (LU) network, using Oyster data.
This research could enable a broader implementation of route choice estimation in large metro networks, especially when researchers can only rely on open data.
