The OR Society’s Criminal Justice Special Interest Group met in London on 27th June. There were four speakers and this is a brief summary of their sessions.
Tim Curtois from Nottingham University spoke about an automated staff scheduling system which had originated as a PhD project, but has subsequently been commercialised through a spin-off company.
The modelling approach allows users to customise rules and priorities and the aim, when running the model, is to minimise the penalties associated with breaching the rules/constraints. A diverse range of rules can be applied, such as full/part-time working, cover/demand, training/skills, personal constraints (like A won’t work with B).
Two problem versions can be modelled: with pre-defined shifts or shifts with no pre-defined times. Tim explained that the model can use two different calculation methods: Exact and Heuristic. The Exact method works best for smaller models but is not so good at arriving at optimal solutions. The Heuristics method is more robust, irrespective of size, but is outperformed by the Exact method on smaller instances.
Tim’s demonstration of the software showed us a couple of different case examples and, in both of them, the software was seen to evaluate around 600,000 roster options per second, with optimal rosters being created in around 30 seconds.
Tom Lidbetter from LSE talked about his work on geographic offender profiling. His PhD project tackles the problem of identifying the most likely location of an offender based on the geographic pattern of offences. The analysis is based on Rossmo’s formula which rates squares on a map for the probability of it containing the location of the offender. He is applying game theory where the players are “the offender” and “the police” who try to guess the location of the offender. Tom also discussed the interesting situation of the “smart” offender who might try to out-wit the police and what their strategy might be for selecting locations to commit a crime.
Mark Thurstain-Goodwin of Geofutures talked about how his organisation brings together geographic information and, for example, census data to highlight interesting patterns and insights that can be of use to policy-makers. He showed some examples of town centre boundaries in the Bristol area and explained how these might be impacted by new supermarket developments. He explained how there is a wealth of data available from the ONS (Output Area Classification data) that could be included in his geographic models. Geofutures delivers GIS insight to any enterprise, anywhere. Their products, available ready to customise, deliver the decision-making intelligence of GIS through a standard web browser.
Finally, I ran a short interactive session called “Processes – what should you measure?”. You’ll find the slides here.