Predicting the Risks of Street Violent Crimes using Agent-based Modeling

Criminological research has shown that most violent crimes are committed by a fraction of offenders, and a small number of geographic locations account for the majority of observed crime patterns. In policing, focused deterrence strategies are widely implemented to predict and prevent the crime occurrence. To facilitate these strategies, we develop an agent-based model to predict risky behaviors of police-focused high-risk individuals by capturing the complex dynamics generated by constant interactions within the autonomous agents themselves and between the agents and the environment. We incorporate both the criminological theories and the expert knowledge into the model and make full use of the public data, such as GIS data, census data, and official crime data. We use Hampton, VA, USA as an example to demonstrate that the model is able to predict high-risk individuals’ likelihood of being involved in street violent crimes and their patterns. This study suggests the model has great potentials in improving the effectiveness of focused deterrence strategies.