Simulating Offender Movement for Crime Prediction

• In Event: Big Data and Crime Prediction

Abstract

Simulating offender behavior in a virtual environment is a new way of gaining insights into crime, with the possibility to forecast future crime developments. While research in Computational Criminology has shown that agent-based modeling offers a potential to improve existing crime prediction tools by modeling dynamic and individual level-behavior, current research using real environmental data is scarce and mainly focused on the movement of known offenders, using fixed locations or locations known by the offender (e.g. home address). In contrast, we examine the basic rules needed to simulate general offender movement along a road network. The rules guiding the offender movement are inferred using human dynamics data and environmental characteristics of the region. Routine Activity Theory and Crime Pattern Theory inform the usage of the data, while historical crime data for the same location is included for performance evaluation.

Authors

• Raquel Rosés, ETH Zürich, Switzerland

• Cristina Kadar, ETH Zurich, Switzerland

• Irena Pletikosa, ETH Zurich, Switzerland