Modelling the effects of social networks on activity and travel behaviour

Abstract

Activity-based models of transport demand are increasingly used by governments, engineering firms and consultants to predict the impact of various design and planning decisions on travel and consequently on noise emissions, energy consumption, accessibility and other performance indicators. In this context, non-discretionary activities, such as work and school, can be relatively easily explained by the traveller’s sociodemographic characteristics and generalised travel costs. However, participation in, and scheduling of, discretionary and joint activities are not so easily redicted. Understanding the social network that lies on top of the spatial network could lead to better prediction of social activity schedules and better forecasts of travel patterns for joint activities. Existing models of activity-travel behaviour do not consider joint activities in detail, except within households to a limited extent. A recent attempt developed at ETH Zurich to incorporate social networks in a single-day optimisation scheduling model did not model joint activities as such, rather rewarding individuals for scheduling activities at the same location and at the same time as their friends. Realistic social networks were also not incorporated. The aim of this thesis is to contribute to this rapidly expanding field by developing a simulation of activity and travel behaviour incorporating social processes and joint activities to investigate the effects on activity and travel behaviour over a simulated period of weeks. The model developed is intended as a proof-of-concept. In order to achieve this aim, an agent-based simulation was designed, implemented in Java, and calibrated and partly verified with real-world data. The model generates activities on a daily basis, including the time of day and duration of the activity. An interaction protocol has been developed to model the activity decision process. Data collected in Eindhoven on social and joint activities and social networks has been used for calibration and verification. Alongside the model development, several issues are addressed, such as exploring which parameters are useful and their effects, the data required for the validation of agent-based travel behaviour models, and whether the addition of social networks to models of this type makes adifference. Sensitivity testing was undertaken to explore the effects of parameters, which was applied to increasingly more complex versions of the model (starting from one day of outputs with no interactions between individuals and finishing with full interactions over many days). This showed that the model performed as expected when certain parameters were altered. Due to the components included in the model, scenarios of interest to policy makers (such as changes in population, land-use changes, and changes in institutional contexts) can be explored. Altering the structure of the in- put social networks and the interaction protocols showed that these inputs do have a difference on the outputs of the model. As a result, these elements of the model require data collection on the social network structure and the decision processes for each local instantiation. Two more "traditional" transport planning policy scenarios, an increase in free time and an increase in travel cost, showed that the model performs as expected for these scenarios. It is shown that the use of agent-based modelling is useful in permitting the incorporation of social networks. The social network can have a significant impact on model results and therefore the decisions made by planners and stakeholders. The model can be extended further in several different directions as new theories are developed and data sets are collected.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.