Machine Learning based criminal short listing using Modus Operandi features

One of the most challenging problems faced by crime analysts is identifying sets of crimes committed by the same individual or group. Amount of criminal records piling up daily has made it cumbersome to manually process connections between crimes. These Crime series´ possess certain attributes that are characteristic of the criminal(s) involved in them, which are useful in defining their modus operandi (MO). After a careful study in the grave crime category of House breaking and Theft in Sri Lanka, we have identified certain MO attributes which we have used to collect from past crime scene data from police records. Then we have explored whether it is possible to group suspects who have similar MO patterns through a machine learning approach and give a short list for a new crime from the existing data. The evaluation of the research presented an accuracy above 75% which proved that Machine Learning is capable of short listing criminals based on their Modus Operandi features.