Public Release: 

And here's the local crime forecast...

New Scientist

COMPUTER forecasts that predict where and when crimes will happen by analysing past patterns should help police channel resources where they are needed most. The technique, now under trial in the US, could be available for routine use within a year.

Police "crime maps" that identify areas prone to illegal activity are commonplace. But a more powerful tool for forecasting crime is emerging from a huge electronic database of 6 million crimes recorded over the past 10 years in Pittsburgh, Pennsylvania, and Rochester, New York.

A team from Carnegie Mellon University in Pittsburgh analysed the data in two ways. The first is a statistical analysis that spots broad trends. This allowed the researchers to quantify, for the first time, the rules of thumb that police officers often learn from experience, such as how much thefts increase in the run-up to Christmas.

Wilpen Gorr, who is an expert on information systems and forecasting and heads the team, also looked for more subtle and unexpected trends. He compiled a list of leading indicators- minor offences such as vandalism and trespassing that crime analysts believe precede more serious crimes. He then used a neural network to make connections that link changes in the pattern of these minor offences with changes in levels of more serious crimes. He found that if a minor crime such as vandalism increases during one month, it can indicate that there will be an increase in serious property crimes such as burglary in the following month.

The team then built a computer model that would reveal whether crime rates reflected these patterns. So far they have tested it using data from a 72-month period in Pittsburgh and Rochester. Their model succeeded in forecasting crime rates with an error of between 10 and 20 per cent in areas as small as individual police beats of only 2 to 3 square kilometres. The accuracy dropped when there were sudden changes in the leading indicators, but the team's rate of success at predicting jumps in crime rates still hovered at about 50 per cent.

According to Don Brown, an engineer at the University of Virginia in Charlottesville who studies crime, the system would be simple to implement and could be distributed to police departments without special training. "You would have to tune it to a particular area," he says.

Gorr and his team plan to do a final test of their program in Rochester this month. They will then start working on a version for general distribution, which they hope to finish within a year.

But you're unlikely to see crime forecasts after the evening news. Brown says that revealing the forecast could reduce its effectiveness. "If people can change the outcome by changing their behaviours, they will."


Author: Emily Singer

New Scientist issue: 16th August 2003


"These articles are posted on this site to give advance access to other authorised media who may wish to quote extracts as part of fair dealing with this copyrighted material. Full attribution is required, and if publishing online a link to is also required. Advance permission is required before any and every reproduction of each article in full - please contact Please note that all material is copyright of Reed Business Information Limited and we reserve the right to take such action as we consider appropriate to protect such copyright."

UK CONTACT - Claire Bowles, New Scientist Press Office, London:
Tel: 44-207-331-2751 or email
US CONTACT - Michelle Soucy, New Scientist Boston Office:
Tel: 617-558-4939 or email

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.