A group of researchers have used the "mobile network structure" to aggregate "human behavioural data" and predict if a certain area of London will be a crime hotspot - and 70 per cent of the time, they're correct.
In a recently released paper, the group identified that criminal profiling is not as effective as policing known microcosms (such as streets or roads) of criminal activity, and sought to use data to help police services operate more efficiently.
The study concluded that using human behavioural data significantly improved the researchers' prediction accuracy, compared with just using census data (economic data, demographics, population, and so on). The paper states: "Our experimental results show that the static nature of these variables makes them less useful in predicting crime levels of a given area when compared with less detailed but daily information about the types of people present in a same area throughout the day."
Also highlighted by the paper was the out-dated method of predicting criminal hotspots using "background historical knowledge about crime events... or wide description of areas using socio-economic and demographic indicators." It also found that using anonymous mobile data more accurately predicted areas of high crime due to the dynamic and up-to-date information.
Although a 70 per cent accuracy rating isn't enough to begin enacting policy change, the report does highlight how government institutions are failing to adopt and utilise new technology.