Two models for semi-supervised terrorist group detection

Abstract

Since discovery of organization structure of offender groups leads the investigation to terrorist cells or organized crime groups, detecting covert networks from crime data are important to crime investigation. Two models, GDM and OGDM, which are based on another representation model-OGRM are developed and tested on nine terrorist groups. GDM, which is basically depending on police arrest data and caught together information and OGDM, which uses a feature matching on year-wise offender components from arrest and demographics data, performed well on terrorist groups, but OGDM produced high precision with low recall values. OGDM uses a terror crime modus operandi ontology which enabled matching of similar crimes. © 2009 Springer Vienna.