Technological developments, such as the rise in GPS enabled devices and Web 2.0 technologies have created social transformations in how we connect and share information through the mass uptake of smart phones and social media platforms. This new generation of mobile technologies work as individual sensors capturing data on a wide range of human behaviours that have been previously hidden. These include data on individual movement, preferences and opinions. Understanding these behaviours is crucial if we are to create a joined up approach to simulating how cities breath and grow.
However, considerable work is required in adapting and developing new technologies from machine learning to extract behaviours which can be embedded into cutting-edge modelling techniques. Creating this bridge between ‘big’ data representing the ‘real’ world, and simulations producing alternative versions of reality is of value to both academics and policymakers looking to develop new solutions to many of the challenges that today’s cities face. To do this we need to understand how factors within the “Social City” (the impact of individual movements and decisions) play out every day in the “Smart City” (data collected from fixed sensors on for example, traffic counts, air pollution or movements of populations).
Explaining the science
Cities are complex systems. Understanding how individuals move around and use urban spaces is a challenging area. Combining sensor data (footfall, traffic) with social data (social media) allows researchers to understand who, where and why individuals are in cities. More importantly, it also allows researchers to construct simulations (using agent-based models or ABMs) that replicate how individuals will react under a number of scenarios.
Whilst agent-based models have become a very popular tool, creating accurate behavioural rule sets is difficult due to a lack of individual level data. This data is now available, but considerable work needs to be undertaken to link new and traditional data sets and then to analyse this data for patterns and relationships between individuals that will inform these models.
Understanding the relationships between individuals is a crucial factor to capture to build into accurate ABMs of city applications. This research will use causal inference modelling (from statistical epidemiology) to quantify these relationships and build them into ABMs.
Finally, this research will examine approaches for understanding and quantifying the uncertainty associated with ABMs. The goal in this area is to examine approaches from the physical sciences that will allow social scientists to provide a level of confidence with their simulation outputs (i.e. to understand where the model under and over predicts).
This research has the following aims:
- Using data wrangling to bring together social and smart city data sets.
- The exploration of how machine learning inspired tools can be used to recognise emergent patterns and processes within geographical and social systems from these new data sets.
- Examining methods to understand relationships between individuals
- Quantification of uncertainty associated with simulations
- Building more robust agent-based models to address important policy scenarios associated with cities.
This work is of interest to any sector or industry that is interested in simulating and understanding how individuals move and interact in urban spaces. In particular, understanding the relationships in populations and the impact on these relationships with the introduction of a new policy (e.g. free transport). This work will also create methods for handling and quantifying uncertainty within simulations.
Ultimately this research could be useful for any stakeholders and industry who are interested in better understanding how cities work, quantifying uncertainty and understanding the implications of implementing different scenarios and policy interventions.