Forecasting crime for strategic crime analysis

Dr Matt Ashby
UCL Department of Security and Crime Science

What forecasting is not

Predictive policing

streets

hours/days into future

Horizon
scanning

cities/countries

(many) years into future

L van Woensel & D Vrščaj. 2015. Towards Scientific Foresight in the European Parliament. Brussels: European Parliament.

Predictive policing

streets

hours/days into future

Horizon
scanning

cities/countries

(many) years into future

Crime
forecasting

cities/districts

weeks/months/(a few) years into future

How forecasting can be useful to policing

Scenario 1

A police chief preparing a budget request wants to know how many crimes are likely to occur in a city each month for the next three years

Scenario 2

A commander in charge of setting officer schedules wants to know how many crimes are likely to occur each day for the next three months.

Scenario 3

The commander in charge of detectives wants to know how many serious assaults are likely to occur in a district each month for the next year.

Forecasting basics

Time series

Time series = trend +

Time series = trend + season +

Time series = trend + season + other + residual

“other” = exogenous variables

  • public holidays
  • sporting events, festivals, etc.
  • disasters and civil emergencies
  • etc.

But, crime analysts …

  • must master lots of skills
  • have limited time
  • have limited software access

⇒ semi-automatic forecasting

Forecasting methods

But …

  • Austin is just one city
  • We have only tested one scenario

Conclusion

  • Reasonably accurate forecasting is possible in some circumstances
  • The combined method is typically the most accurate
  • Some specific methods (e.g. FASSTER) are wildly inaccurate

These slides: lesscrime.info/talk/forecasting-dhpol/

Questions: matthew.ashby@ucl.ac.uk