In the realm of criminal justice, recidivism pertains to individuals exhibiting persistent patterns of reoffending. This subset of repeat offenders poses a significant burden on the policing system due to its recurring nature. An ability to predict recidivism holds the potential to empower law enforcement with strategic interventions aimed at disrupting this cycle. This study delves into the Recidivism Forecasting Challenge presented by the National Institute of Justice in the Summer of 2021. Two years of data from the State of Georgia and two performance metrics, Brier score and fair and accurate, are used. The data involved a mix of both qualitative and quantitative input variables with potentially complex relationships with the probability of recidivism. In this research, application of a classification model, potential data challenges and scopes are discussed.
Haque et al. (2024) studied this question.