Capturing and modelling complex decision-making in the context of travel, time use and social interactions

The field of choice modelling is evolving rapidly, with ever more complex representations of heterogeneity and a growing interest in discrete-continuous models instead of just discrete choice. Rapid developments are also taking
place in terms of new data sources, with a growing revival of revealed preference data instead of stated preference, notably in terms of ubiquitous data and longitudinal surveys. Against this backdrop of developments in choice modelling, there is also a growing recognition of the role of the social environment on behaviour, through the effect of close and far social network members, with social interactions increasingly relying on digital communication. This thesis makes contributions in all of these areas, and differently
from previous research, which has often focussed on just one, attempts to jointly explore multiple dimensions. Relying exclusively on revealed preference data, the work provides important insights into how social networks
evolve over time, how people within a network interact with each other, and how there are links between a person's social network and his/her activity
scheduling. The empirical results provide valuable insights for researchers
not just in transport, but also in other fields. The behaviour modelled in this thesis is complex, with heterogeneity in different dimensions, in terms of preferences as well as behavioural processes, at the person level as well as at the individual choice level. At the same time, many of the choices are not mutually exclusive and have a continuous dimension too. The work
makes a number of modelling advances, in terms of facilitating the recovery of heterogeneity and in putting forward solutions that allow for a computationally tractable representation of correlation and complementarity in discrete continuous choice, both in estimation and forecasting. Ever more complex representations of behaviour rely on rich data. This thesis highlights the benefits of detailed real world datasets in this context, and provides important insights into the limitations of existing data sources. The thesis closes by introducing a rich unified data collection that can be used for choice modelling applications in different fields.