Using vast amounts of unstructured data from a massive dump of e-mail published by Wikileaks, a group of researchers has built a computational model that can predict where violence will occur in Afghanistan as well as the level of its intensity. The group published its results in the online Proceedings of the National Academy of Sciences in mid-July.
For years now predictive analytics has been used to anticipate where in a city a police department should deploy officers in order to prevent violent and property crime. However, in those cases, as far as I know, the information used in the mathematical models has been structured, based on crime data gathered in a database. This is the first time I’ve encountered analysts using unstructured data to predict the location and level of violence in a war zone.
Perhaps equally impressive is that the work reportedly was done in a single night.
The benefits of such a model will be obvious to military commanders, particularly in Afghanistan, where it can be used to save civilian lives and help protect troops in the field while clamping down on an enemy. But the benefits should also be apparent to leaders of almost any kind of organization. Given the right kind of unstructured information and the right analytics model, valuable insight can be gleaned. As the Wikileaks e-mails indicate, however, sources of the “right” information may not always be obvious. It may take some creative thinking. But the effort to find and mine that data may be well worth it.
Image credit: Reuters/Pichi ChuangComments