27 February 2014

One of the most important objectives for the Ministry of Justice (MoJ) is to analyse vast amounts of crime and offender data to identify hidden patterns of behaviour and prevent reoffending.

The principal risk assessment and management tool used by the National Offender Management Service is the Offender Assessment System (OASys). This is used by around 140 prisons and 35 probation trusts. Information from all completed OASys assessments is collated centrally within the OASys Data Evaluation and Analysis Team (O-DEAT) database, which records information from millions of prisoner assessments. It includes data on individual offender circumstances such as accommodation, education, lifestyle, financial management and income, relationships, etc.

The MoJ identified that merging OASys data with other types of data held in its files would be essential in order to gain a clearer and more comprehensive view of the offender. But with almost four million records on file, it wasn’t feasible to trawl through the data manually in an attempt to identify those factors that may mean a prisoner is likely to reoffend.

The ministry turned to IBM’s SPSS Statistics and SPSS Modeler Premium software to analyse the files. For example, the technology helps identify whether offenders with specific problems such as drug and alcohol misuse are more likely to reoffend than other prisoners.

As a result, forecasts about recidivism have substantially improved. In the case of violent crime, forecast accuracy has gone up from 68 to 74 per cent, while predictions for general offences have improved from 76 to 80 per cent.

The MoJ says: “IBM SPSS predictive analytics technology enables us to advise on preventative measures, such as appropriate programmes addressing offender behaviour, before a prisoner’s release.”

In addition, O-DEAT deals with ad hoc requests for data analysis, providing vital management information to policy colleagues, management, and practitioners in the field. IBM says the Modeler data mining workbench has also helped to speed up and simplify the process of data extraction and manipulation to meet such requests.