Avon and Somerset Constabulary has developed a predictive analytics engine to tackle domestic violence and child abuse.
A report last year on improving the police response to domestic abuse from Her Majesty’s Inspectorate of Constabulary (HMIC) identified a number of failings at Avon and Somerset.
The report stated: "The public can be less confident of the constabulary’s organisation and arrangements to manage the safety of all victims of domestic abuse.
"The management of high risk victims is generally good, but medium and standard-risk cases are not consistent across the constabulary and some parts of the procedural management of cases is disorganised."
The report found the constabulary did not have robust quality-assurance checks or qualitative auditing across many areas of its procedures, which would provide an insight into the effectiveness of police action or the quality of service victims receive.
"In some cases, staff do not always get the full picture of risk from the victim or the police databases," according to HMIC.
At the IBM event, Sean Price, head of performance at Avon and Somerset constabulary, demonstrated how the constabulary had developed a proof of concept for predictive analytics, which was being rolled out to identify and protect victims of domestic violence and child abuse.
More articles on predictive analytics
- Do predictive analytics projects require data scientists?
- Gazing into the future of predictive analytics ROI
- Predictive analytics keeps business leaders' fingers on the pulse
"We believe predictive analytics is the solution. Our analysts can deliver a higher level of value to the organisation by triangulating technique, which allows officer to make better decisions," he said.
The system, developed at Avon and Somerset constabulary, uses historical crime data, along with textural and sentiment analysis combined with additional databases and open-source information, to create a statistical model that can predict an individual's behaviour and risk.
Built on IBM Predictive Analytics, the system produces a risk score, allowing officers to identify potential victims before they are harmed.
"We can see which domestic situations are likely to escalate," said Price. "We are changing our transactional and volumetric approach to risk and providing a visible safety net, which focusses people's minds in the right areas and helps us identify potentially serious cases we have not yet identified."
He said the statistical model used by the constabulary is able to learn what makes people a risk.
"You then flush through all the data in our organisation and show the escalation of risk," said Price.
He added that the constabulary has partnered with the University of West England to evaluate the use of predictive analytics compared with traditional policing.