Predictive technology gives Kent Police visible presence in community policing

Kent Police showcased the power of its latest predictive policing technology during a day of action that led to almost 20 arrests for offences including assault, possession of drugs and carrying a knife.

Known as ‘PredPol’, the predictive policing software is an intelligence-led computer program that uses an algorithm to pin-point officers to zones (boxes) where crime is most likely to occur. The program was developed in Los Angeles and utilises advanced analytics to read the crime data against patterns of criminal behaviour to allow the force to get the best use of its resources. Officers receive updates from the system twice a day and then decide where to patrol.

The crime-prediction boxes are the result of the same calculations used to predict earthquakes.

Analysts have described predictive policing as a natural progression from crime mapping.

Kent Police Chief Constable Ian Learmonth, who initially trialled the software in December, said: “We are the first police force in the UK to fully embrace predictive policing and the results so far have been encouraging.

“When we piloted PredPol in North Kent we saw a six per cent reduction in street violence.”

The day of action on August 14 saw 95 officers dedicated to patrolling the predictive policing zones. This resulted in 1,219 visits to zones and 220 hours of patrolling.

“There were no additional costs to the force – this was business as usual – but we believe it led to a ten per cent reduction in crime, particularly violent crime, over the 24-hour period and a reduction in calls to our control centre,” said Mr Learmonth.

He added that predictive policing was not just about making arrests. “It is aimed at reducing crime, as well as increasing engagement with local residents and businesses. By having a visible presence in these areas we are also able to provide assurances to communities,” said Mr Learmonth.

“Officers still go on patrol as normal and the boxes give them an idea of where to concentrate their efforts. If nothing happens while they are patrolling it is still successful; we may have prevented a crime from being committed. Despite the fact this utilises modern technology, the success of this approach lies within good old-fashioned policing – officers out in the community, policing what they see.

“The public wants to see visible policing on the streets; this allows us to do that.”

Kent Police is now working closely with forces across Europe which have shown an interest in adopting this intelligence-led approach to policing.

In the next day of action planned for later this year, officers will target the night-time economy.

Kent’s initiative took place at the same time as similar action by US forces the Los Angeles Police Department (LAPD) and the Santa Cruz Police Department, California, where the software was first tested in 2011. Crime there dropped by 27 per cent and 12 per cent respectively in the first six months of its use.

PredPol began as an academic project in the US seven years ago, involving mathematicians, social scientists and led by Jeff Brantingham, an anthropology professor at the University of California whose goal was to understand how crime hotspots occur.

The group found that by using a mathematical algorithm similar to that for predicting earthquake aftershocks, they could predict where crimes would happen.

Professor Brantingham explained that when there is an earthquake you learn a lot about how it produces aftershocks, which in general occur near in time and place to a main shock. Crime is very much the same, he said: “If you look at crime hotspots they appear quickly and spread around. When a burglar breaks into a house it makes the chance of crime occurring close by much more likely.”

The PredPol software maps years of past crime data and is updated daily with the location, time and type of crime committed. From this it creates prediction boxes of precise 500sq ft zones which are listed in priority order as to where crimes are most likely to occur, which is then delivered to the