Algorithms is useful in so many ways that it can even be used to fight crimes. Several police departments are now tapping the potential of using algorithm in their fight against crimes. Basically, heaps of statistical data about past criminal events in a certain location are used to extrapolate when the next crime will happen in the same area.
Researcher Mirco Musolesi however would like to put a twist into his new study by focusing instead on where the next crime will happen, instead of the when. His method is a little controversial because he is not using old crime data, but instead by following the movements of potential criminals and their movements via their mobile phones.
His idea was borne out of his research into what he calls mobility patterns, a paper he published at the University of Birmingham in the U.K. He won Nokia’s Mobile Data challenge by following and predicting the movements of 25 volunteers in a town in Switzerland. He tapped a few number of information like GPS data, texting and calling history, and telephone numbers to track the movements of subjects. His algorithm sometimes was able to predict the target destination of volunteers within 20 meters.
Interestingly, the algorithm was as precise as demonstrated above only when it puts into the consideration the data of the friends of a volunteer. When the algorithm is merely using the data of a volunteer, it could only predict the future coordinates to within about 1000 meters.
The obvious downside of the study is the moment a criminal turns off its smartphone or any tracking apps and GPS. Musolesi insists though that this should not stop his algorithm from tracking potential lawbreakers because his metrics can also be applied if the police will use cell phone tower information.
Cell towers are not as precise as GPS data though. However, if network operators could provide information (after a court order) authorities can track a potential criminal’s movement by use of triangulation from the nearest base station. The researcher said that the crucial key is to look for patterns of movements, so timely intervention can be done after the algorithm suggests future movement to an unusual area.
The published study says that authorities today already have more than enough information to determine the position of a suspect if the need arise, but it’s still a matter of making use of existing data to predict where the crime will take place. The algorithm’s main function then is to give time for authorities to prepare themselves.
Musolesi hopes his work will be applied to law enforcement by testing how the algorithm predicts crime locations. He also suggests to use it on anonymous people on bail in the United Kingdom who are tagged electronically.
A relatively similar algorithm is being used by Facebook to identify suspicious users. In one instance, Facebook was able to tip-off the police about an arranged date between a man and a 13-year old girl after the social network site tracked frequently used words like “sex” and “date”.