to add, there is also training data available in form of people who have committed shootings or have any crime record. its then just about feature selection that could prove some causation or correlation