Applying Unsupervised Machine Learning to Counterterrorism

Abstract

To advance the agenda in counterterrorism, this work demonstrates how analysts can combine unsupervised machine learning, exploratory data analysis, and statistical tests to discover features associated with different terrorist motives. A new empirical text mining method created a “motive” field in the Global Terrorism Database to enable associative relationship mining among features that characterize terrorist events. The methodology incorporated K-means co-clustering, three methods of non-linear projection, and two spatial association tests to reveal statistically significant relationships between terrorist motives, tactics, and targets. Planners and investigators can replicate the approach to distill knowledge from big datasets to help advance the state of the art in counterterrorism.