Evaluating class imbalance and asymmetric costs using machine learning

Abstract

The current study will evaluate the use of machine learning as a form of risk assessment as it fits within the risk-need-responsivity framework. Specifically, the risk of violent reconviction will attempt to be predicted by multiple machine learning algorithms. As violent reconviction has significant class imbalance, as well as asymmetric error cost, methodologies accounting for these potentially problematic situations will be evaluated. While machine learning has been shown as an improvement over traditional assessment, more research is necessary to determine the most effective practices when applying its specialized methodologies. Analysis of the techniques used as treatment for class imbalance and asymmetric cost has not been researched on actual criminal justice data, leading to a gap in the scientific literature necessary to evaluate their genuine performance when applied.