The US military is developing a 'big data' tool aimed at preventing military personnel from committing violent crimes by predicting which soldiers are most likely to offend.
Individuals identified as 'high risk' can then be offered counselling.
Using historical data on hundreds of thousands of soldiers, researchers from Harvard Medical School have developed an algorithm that can accurately identified some soldiers at high risk of carrying out violent crimes, according to a report in the Los Angeles Times.
The programme looked at the military records of more than 975,000 who served between 2004 and 2009.
In that period 5,771 soldiers committed violent crimes such as murder and robbery.
The researchers used computer programmes to identify certain characteristics displayed by high-risk individuals.
The "key predictors" identified by the researchers in an article in the journal of Psychological Medicine, included disadvantaged social backgrounds, prior offences, early career stage and prior treatment for mental health disorders.
Among male soldiers, the researchers found the highest risk group, which constituted about 5 percent of the total, accounted for 36 percent of crimes by men.
Among female soldiers, the highest risk group was responsible for 33 percent of crimes by women.
Researchers tested the model on a sample of more than 43,000 soldiers who served between 2011 and 2013, and found the most at-risk category were responsible for 51% of violent crimes committed by those soldiers.
The study, which has yet to be employed by the military, is an extension of earlier efforts to use data sets to identify soldiers at risk of suicide.
