Find a SECReT supervisor
prism apply now

Prediction of crime patterns emerging from simulated search trajectories of individual offenders

22 February 2012

Lucy Burton

The behaviour of criminals searching for targets has close analogies with animal foraging. This study used agent-based modeling to simulate offender trajectories assuming two types of foraging strategy. Trajectories were modelled so that crime pattern predictions could be generated under various hypothetical conditions. Crime patterns were represented in terms of inter-crime distances (where crimes are committed sequentially without the offender returning home) and journey-to-crime distances (where the offender returns home after each crime).

In a range of experimental conditions, it was shown that both these distances follow a long-tailed distribution resembling an inverse power law, modified by a truncation at smaller values. However, at low values there was deviation from this behaviour. Statistical analyses showed that log-normal and Weibull distributions provide a reasonable fit. Model equations derived from the generated crime patterns were developed and offered as a potential method for analysing empirical crime data.

In addition, it was shown how such equations could be applied to devise a search algorithm that may be used for geographic profiling of crime. Data on domestic burglary in London (supplied by the Metropolitan Police Service) were analysed with the aims i) of assessing similarities between animal foraging and crime patterns and ii) estimating the extent to which real crime patterns may be used as a validation of those predicted from agent-based models.