UCL News


Crime sleuths crack down on home break-ins using predictive maps

5 May 2004

A novel method of predicting where home break-ins will occur that is 30 per cent more accurate than current crime mapping techniques has been developed by University College London crime sleuths.

Published online today in the British Journal of Criminology, researchers from UCL's Jill Dando Institute of Crime Science show it's possible to develop more effective crime hotspots maps by treating the spread of crime like a disease.

Previous work by the team has shown that the risk of having your home broken into follows a similar pattern to the way a disease is spread between people. Results indicate properties within 400 metres of a burgled home, and particularly those on the same side of the street, are at an elevated risk for up to two months after the initial event.

The new study confirms the technique, which generates what is known as a 'prospective hotspot map', is more effective than traditional crime mapping methods that identify high crime locations based on incidents that have already occurred.

Dr Shane Johnson of UCL's Jill Dando Institute of Crime explains:

"Traditional policing and crime prevention following a break-in focuses on making the victim and their property secure. But burglaries often occur in clusters because house layouts, exit routes and the kind of 'pickings' can be similar.

"Crime maps can help predict where crime hotspots might occur but currently these are generated using historic area crime rates. Prospective hotspot maps are more dynamic. This means calculations of where hotspots occur are constantly adjusted, or evolve, to reflect any new incidents that occur.

"Crime doesn't stay in the same place. Using this approach we can predict the new patch a perpetrator is moving into and identify the households that are most at risk. Homeowners can then be informed so they can be on their guard and police patrols can be targeted to make more efficient use of resources."

Hotspot maps are generated by laying a grid over a conventional map of the area under investigation and applying algorithms to describe which areas have previously been hit by crime. Prospective hotspot maps differ because they use algorithms that consider not only where but also when incidents occur to calculate the risk factor for each cell in the grid. The maps are more accurate than other methods used to pinpoint future crime hotspots because calculations of hotspots are based on the distance (400 metres) and time over which the researchers have discovered the risk of burglary is communicable.

"One problem with previous crime mapping techniques has been the lack of robust evaluation to test how good they are at predicting the location of future crimes," says Dr Kate Bowers of the Jill Dando Institute of Crime Science.

To test whether prospective hotspot maps are more effective than other methods of generating crime maps, the team designed a new system of map evaluation.

A number of measures of hotspot accuracy were considered, including:

* numbers of new crimes that are successfully predicted in the hotspot area
* size of the hotspot area
* numbers of crimes successfully predicted per kilometre squared
* numbers of different hotspots areas produced by the technique
* calculation of the ratio of hotspots to the size area under study

Hotspot maps were generated using the new prospective method and two traditional hotspot methods with data on the number of break-ins in the county of Merseyside. Accuracy was tested by using two months data and then comparing the predictions generated with what happened in the next two months.

Results show that prospective mapping captured between 62 to 80 per cent of burglaries and retrospective mapping captured only 46 per cent of incidents. On average prospective hotspot maps were shown to be 30 per cent more accurate.

Dr Johnson added: "We have to stay one step ahead of the criminal fraternity and while we're not at the stage of science fiction films like Minority Report anything that can help make us that bit safer must be investigated thoroughly."

Notes to editor

For more information please contact Judith Moore on +44 (0)20 7679 7678 or mobile: +44 (0)77333 07596.

Journal: British Journal of Criminology
Title: 'Prospective hotspotting: the future of crime mapping?'
Authors: Kate Bowers (1), Shane Johnson (1) and Ken Pease (1)
(1) Jill Dando Institute of Crime Science, University College London, EC1V 9HL

About the Jill Dando Institute of Crime Science

The Jill Dando Institute of Crime Science was established on the second anniversary of her death. Based within the School of Public Policy at University College London, it has built an international reputation in the area of crime science in its three years of existence. The Institute's broad aims are to: change the way in which crime and its control are thought about; apply scientific techniques to crime prevention and detection; improve quality of life through research, teaching and influencing public policy; and draw together reliable facts and research focused on the problems of crime. See the website at: http://www.jdi.ucl.ac.uk/