Introduction

Understanding how police supply responds to changing demand is crucial with ever greater pressures on the police force. Computational models offer a potential means for replicating the highly complex interdependencies that exist in police supply. This project generated a proof-of-concept method for simulating police demand on a day-by-day basis and an agent based model of police supply.

Explaining the science

The 21st century has seen policing agencies become involved in an increasingly diverse range of roles, often while managing relatively restricted resources. Consequently, a key priority for applied policing relates to better understanding and anticipating changes in short-, medium- and long-term demand. As a result, models of strategic police resourcing are required to allow police agencies to better optimise the allocation of existing resources in response to demand. Yet, understanding these resourcing problems is a non-trivial task. An array of factors, both internal and external to police organisations, influence demand and resourcing dynamics. Moreover, these factors are often highly interdependent (meaning that all choices have opportunity costs) and difficult to model using traditional analytical techniques. 

In response to these challenges, this project looks at how data-driven agent-based models might be applied to better understand the dynamics of police demand and resourcing. The demand model developed works by looking at historic crime data in a specific geographical district, at a specific day of the year, for a specific crime type, building a Poisson distribution and sampling randomly from that distribution. This gives a hypothetical number of counts of a specific crime type, in a specific geographical area in a specific day of the year.

The model of supply is agent-based, where individual 'agents' of police resource (police officers) are allocated crime reports from the demand model at a regular time interval. These jobs are distributed between agents by a simple queue and agents then work on that job for a specific amount of time in hours determined using crime severity scores from the ONS. Multiple agents can be allocated to the same case based on the crime severity score. Agents are also rostered on and off shifts to replicate the real working of police forces. This begins to explain how backlogs of certain crimes can occur and how changes in the supply (number of agents) can affect the day-to-day operation of a police force.

Project aims

This work aimed to show that computational models offer a robust, transparent and extendable toolkit for modelling how police forces respond to changes in demand. It looked to build a proof-of-concept system to specifically model police demand in the context of crime and begin to model how police forces respond to crime. Understanding whether models could be built to replicate this highly-complex real world system are incredibly important given the ethical and logistical issues attempting experiments on existing real world systems. Therefore, simulations offer the possibility to replicate real-world systems in a safe and effective manner which can then be used to provide insight into the real-world system. 

Project objectives:

  1. Explore potential agent-based computing applications in the modelling of police resourcing and demand problems at varying scales of abstraction (i.e. operational, strategic)
  2. Engage with stakeholders and support collaborative systems mapping exercises to inform ongoing model development. 
  3. Explore what types of data may be required / are available to calibrate and validate simulation models.
  4. Explore how other data science methods might be used to most effectively capitalise on these data and inform the broader project.

Applications

Police forces around the UK could benefit from computational models that are able to offer alternative future scenarios around future demand that provide medium to long term strategic insights into how best to recruit and deploy operational supply.

Organisers

Researchers and collaborators

Funders