Using Artificial Intelligence to Model Juvenile Recidivism Patterns

Abstract

Abstract Risk management has had a major positive impact on increasing the effectiveness of probation supervision. However, while methods and procedures for designing and implementing such a system are well known, there is still a lack of utilization among many juvenile courts. Discriminant classification and neural network models were developed to decide the set of classification variables that would significantly differentiate recidivists from non-recidivists. These models correctly differentiated between recidivists and non-recidivists in 63 percent (discriminant) and 99 percent (neural network) of the cases respectively.