Organized Crime Structures in Co-offending Networks

This paper aims at a conceptual foundation for the development of advanced computational methods for analyzing co-offending networks to identify organized crime structures -- i.e., any static or dynamic characteristics of a co-offending network that potentially indicate organized crime or refer to criminal organizations. Specifically, we study networks derived from large real-world crime datasets using social network analysis and data mining techniques. Striving for a coherent and consistent framework to define the problem scope and analysis methods, we propose here a constructive approach that uses mathematical models of crime data and criminal activity as underlying semantic foundation. Organized crime has been defined in a variety of ways, although, so far, there is surprisingly little agreement about its meaning -- at least not at a level of detail and precision required for defining this meaning in abstract computational terms.