A team of physicists from Howard University spent their summer at START attempting to decipher patterns in START datasets and develop algorithms that could eventually enable the prediction of terrorist threats.
“Even if such techniques are not able to predict future terrorist attacks with certainty, it is anticipated that such studies will assist the intelligence and homeland security communities in making informed decisions regarding deployment of counterterrorism resources to effectively thwart terror plots prior to their occurrence,” said Prabhakar Misra, Professor of Physics at Howard University.
Led by Misra, Raul Garcia-Sanchez and Daniel Casimir examined START’s Global Terrorism Database and the forthcoming Profiles of Incidents involving CBRN by Non-state actors (POICN) database looking for links between terrorist attacks based on region, terrorist organization, given terrorist events and future attacks.
The advanced computational techniques they utilized, including data mining methods and machine learning, allowed them to envision attacks and the outcomes for a particular target variable.
While some of the techniques they used were similar to methods utilized in their study of physics, the project required them to apply their physical science understanding to the world of social science. They also had to learn how to use new techniques and describe their findings in a way that bridged both fields.
“This experience proved highly rewarding as our mentors at START were both knowledgeable and supportive of our efforts and all of us gained familiarity with the utility of data mining methods as applied to ‘Big Data,’” Misra said.
The summer also left the whole team of researchers with an appetite to conduct more research in this area.
Garcia-Sanchez would like to employ additional approaches to developing algorithms for the GTD that would significantly reduce the time it takes for its machine learning process. Casimir would like to resolve the challenges he had in reconciling two data mining efforts for POICN. Misra is interested in applying and automating machine learning algorithms to identify and handle missing and incomplete data within databases such as the GTD, POICN and Profiles of Individual Radicalization in the United States.
The team was at START thanks to a competitive Department of Homeland Security Summer Research Team Program designed to support faculty and students from Minority Serving Institutions.