Dr Alex Coleman, Dr Daniel Birks, Prof Alison Heppenstall – University of Leeds.
This work was supported by Wave 1 of The UKRI Strategic Priorities Fund under the EPSRC Grant EP/T001569/1, particularly the “Criminal Justice System” theme within that grant & The Alan Turing Institute.
Simulating the thin blue line
Over the last 20 years the demands on the police force have dramatically changed. With the changing nature of crime, greater budget constraints and an emerging role as a service of last resort for the most vulnerable. How police forces plan for changes in these demands is a crucial strategic question. However, demand is highly complex and how police supply responds to demands to complicated by a variety of trade-offs and interdependencies. Nevertheless, computational models offer a potential avenue for simulating these complex systems and could offer insight to help police forces plan for future changes in demand.
The project aimed to build a proof-of-concept approach to modelling police demand and supply. It aimed to develop a toolkit for forecasting future rates of crimes on a daily basis within a specific geography and a preliminary agent based model that could model basic responses by police forces to crime data.
Explaining the science
To model demand we built a simple sampling system that took past crime occurrence data and built a Poisson distribution from which future crime counts were sampled. This was performed for a variety of crime types at specific geographies to build a synthetic year of crime events. In the supply model, we utilised an agent-based model built in NetLogo. This involves creating individual agents that represent police officers, they roster on and off shift over the course of a day and are allocated crime events as they occur from the demand model. Simple rules are set relating to the amount of time and number of agents required to work on particular crimes allowing us to model emerging behaviours from this system.
Over the course of this project we have developed an open source package, the crime_sim_toolkit (https://github.com/Sparrow0hawk/crime_sim_toolkit) which allows for users to simulate crime events for an entire year at a daily level. This modelling system is currently able to predict rates of crime at the aggregate level of Leeds to within 10% of observed crime rates. The data output of the toolkit is passed to a proof-of-concept agent based model that can model how police respond to increases or decreases in specific types of crimes. Both these models will be the basis for a larger piece of research into police demand and supply funded by the Alan Turing Institute that will capture data around other forms of police demand rather than just crime.
Simulations of this kind could provide the basis for strategic decision-support tools that enable police forces to better prepare for changes in demand in the medium and long-term.
Funders / Partners
This project was in partnership with collaborators at UCL and funded by The Alan Turing Institute
Above: A screenshot of the agent based model at work.
This project was undertaken as part of the LIDA Data Scientist Internship Programme.