Mathematicians have developed a machine learning system that, according to reports on the Internet and other open data, can determine the approximate time and place of attacks a week before they are committed: https://argoprep.com/blog/k8/go-math-grade-4-vs-argoprep-grade-4-workbooks/. The description of the development was published by the scientific journal Science Advances.
“Several thousand people die every year from terrorist attacks. Using only publicly available data, we have created a model that can predict many terrorist attacks a week before they are committed in those regions of the world where terrorists are most active,” the researchers write.
In recent years, scientists have begun to actively explore how information is disseminated over the Internet and how such data can be used to obtain information about the character traits and behavior of users of such sites.
For example, four years ago, American sociologists created a system that makes it possible to assess the likelihood of divorce by messages on Facebook, and their Russian colleagues learned to calculate the marital status of users and the gender of a person by their activity on social networks: https://argoprep.com/blog/k8/go-math-grade-5-vs-argoprep/.
In a new work, mathematicians led by Professor of Zhejiang University André Piton wondered whether it is possible to use open sources of information to determine in advance when and where terrorist attacks will occur.
To test this idea, scientists have collected a huge amount of information, including data on terrorist attacks that were carried out by various extremist groups in all corners of the world between 2003 and 2016. By analyzing this data using various machine learning algorithms, mathematicians tried to find indicators by which one could predict further terrorist attacks.
These calculations https://argoprep.com/blog/k8/go-math-grade-6-vs-argoprep/ showed that machine learning systems can indeed make such predictions in those cases when they work with data that comes from those parts of the world where terrorist attacks occur quite often. In such cases, algorithms can predict about half of the incidents in the week before the terrorist attack took place.
With the help of these systems, Python and his colleagues assessed which socio-political factors most strongly influenced the actions of terrorists and their choice of the place of committing terrorist attacks. In particular, it turned out that extremists were most active in those regions of the countries that were characterized by a high contribution to the region’s economy, low population density, and well-developed road surface.
According to scientists, specialized intelligence services and governments of those countries of the world where terrorist attacks occur most often can use the algorithms developed by them to determine where and when subsequent terrorist attacks can occur. This will increase the chances of preventing them and allow the authorities to spend less resources on fighting false attacks, summed up Python and his colleagues.