A Comprehensive Analysis of Discrete Choice Modelling Specifications for Modelling Route and Stop Choice Behaviour of Transit Users

Public transit demand modelling typically considers two main aspects, namely: mode choice and transit assignment. In such a framework of modelling transit users’ behaviour, three important decisions are considered: choice of mode, choice of access to a stop and choice of a route. The literature argues that transit users may evaluate different strategies such as minimising travel time, travel cost or number of transfers while choosing transit stop or route. Furthermore, there are other important aspects regarding the behaviour of users like how to model individual behaviour in making decisions based on the behavioural theories like random utility maximisation (RUM) or random regret minimisation (RRM), and the influence of users’ socioeconomic, demographics, and trip attributes on these decisions. This study addresses all these aspects using different discrete choice modelling formulations to evaluate the appropriateness of these specifications for access stop choice and route choice modelling. In this study, it is found that models based on the RUM theory generally show a better fit to the data which is used in this thesis. However, a hybrid RUM-RRM specification shows better prediction capability. Particularly for the transit route choice models in this study, models with smaller choice set sizes work better than the model based on larger choice sets. However, the prediction capability is better in models with larger choice sets. Another finding of this thesis is that accounting for different strategies in modelling the behaviour of transit users is quite important. This thesis also examines the effectiveness of the widely used simple random sampling technique in different discrete choice formulations in the route choice modelling context and proposes approaches to improve the quality of the choice modelling specifications. The use of the proposed sampling mechanisms can be expanded to other contexts like destination choice, housing search, and route choice, where sampling of alternatives is inevitable. Finally, the findings of this thesis research can provide insightful modelling methods to be used by transport/transit planning professionals to improve the quality of their demand models