Detecting criminal networks: SNA models are compared to proprietary models

Abstract

Criminal networks have been an area of interest for Public Safety and Intelligence Community as well as social network analysis and data mining community. Existing literature shows that offender demographics and crime features are not taken into account to identify their possible links to find out criminal networks. Four crime data specific proprietary group detection models (GDM, OGDM, SoDM, and ComDM) have been developed based on these crime data features. These specific criminal network detection models are compared more common baseline SNA group detection algorithms. It is intended to find out, whether these four crime data specific group detection models can perform better than widely used k-cores and n-clique algorithms. Two datasets which contain various real criminal networks are used as experimental testbeds.