Exploring the Determinants of Crime-Terror Cooperation using Machine Learning
- PDF / 2,052,376 Bytes
- 32 Pages / 439.37 x 666.142 pts Page_size
- 9 Downloads / 224 Views
Exploring the Determinants of Crime‑Terror Cooperation using Machine Learning Julia Semmelbeck1 · Clayton Besaw2
© Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract Objectives This study seeks to further strengthen extant knowledge regarding terrorist group involvement in organized criminal activity through two means. First, it measures a set of environmental and organizational characteristics for a sample of well-known terrorist organizations based on the crime-terror literature. Second, it illustrates the utility of inductive research designs for examining patterns in the criminal behavior of terrorist groups for theory building and the potential risk classification of new terrorist organizations in the future. Methods The authors utilize a random forest classification algorithm to examine three sources of information about a broad set of environmental and organizational factors determined to be of potential importance in predicting when a terrorist organization will engaged in organized criminal behavior. First, it examines out-of-sample accuracy through bootstrap cross-validation estimation. Second, it quantifies the predictive efficacy/importance of each measured factor. Finally, it utilizes partial dependence functions to examine the relational trend between the most important predictive factors and variation in the presence of organized criminal behavior. Results The study finds three results. First, predictive accuracy using readily quantifiable factors about the criminal behavior of terrorist organizations is good but could be improved upon. Second, organizational factors such as group size, ideology and attack behavior out perform environmental factors in terms of predictive performance. Third, it finds that the most important predictor variables have a predominately non-linear relationship with whether the algorithm would classify a group as engaging in organized criminal behavior or not. Conclusions The study finds that theory building should seek to examine temporal variation in the organizational structure of terrorist groups as a fruitful way forward for further understanding when a group is likely to engage in organized criminal behavior. It also suggests that scholars should seek to engage more critically with concepts surrounding the potential non-linear pathways in which groups end up engaging in organized crime. * Clayton Besaw [email protected] Julia Semmelbeck [email protected] 1
University of Mannheim, Mannheim, Germany
2
One Earth Future Foundation, Broomfield, USA
13
Vol.:(0123456789)
Journal of Quantitative Criminology
Finally, the results illustrate the utility of modern machine learning algorithms and inductive research processes for both academic and practitioner needs alike. Especially when dealing with a complex phenomenon with imperfect data. Keywords Terrorism · Organized crime · Machine learning
Introduction Terrorist groups that get involved in organized crime pose salient security challenges to regions across the worl
Data Loading...