The Research Project “Engaging Big Data” aims to turn passively generated streams of geographic and activity data into predictive models of activity and travel behaviour, in order to test the viability of policy and infrastructure decisions before they are implemented, and guide and inform the urban and transport planning process.
The goal is to dramatically improve the turnaround time, and lower the barrier-to-entry for implementing the latest agent-based, activity-oriented approaches to transportation modelling, applied to the open-source agent-based modelling framework, MATSim. Importantly, the methods developed should be applicable in both developing and developed urban contexts. A further goal will be to improve the means of protecting the privacy of people generating the data that drive these models, through machine learning and automated data synthesis.
As a first application, we developed a agent-based public transport simulation that runs primarily based on Smartcard data. Diverse statistical models are applied to extract critical behavioural and operational parameters to set up a Multi-Agent Transport SIMulation (MATSim) model. What sets the model apart from existing approaches is that it accounts for dynamic phenomena such as bus bunching, vehicle overcrowding and congestion. Hence, for the first time, public transport operators and transport planning authorities will be able to evaluate the impact of alternative vehicles types, new service lines and entire new network configurations not only with regards to ridership, but also service reliability, crowdedness and individual customer satisfaction.