S.T. Karamouzis, A. Katsiyannis, and T. Archwamety (USA)
Neural Networks, Forecasting and Prediction, Juvenile Recidivism
In the U.S. juveniles are involved in almost a third of arrests for major crimes. Efforts for successful rehabilitation for delinquent youth have posed a challenge and generated controversy. Nonetheless, identifying factors related to recidivism may help researchers and practitioners develop more effective preventative and intervention programs for adolescents. Despite intense efforts to identify factors that discriminate recidivists from non-recidivists, prediction models generally account for 20% or less of the variance in recidivism. This article presents the development, training, and testing of an Artificial Neural Network for predicting juvenile recidivism. The network was developed as a three-layered perceptron and was trained using the backpropagation principles. For training and testing various experiments were executed. In these experiments, a sample of 166 profiles of juveniles was used. The sample was divided into two sets. The first set of 120 profiles was used for training and the remaining 46 profiles were used for testing. The predictability rate for the training and test sets were 100% and 74%, respectively.
Important Links:
Go Back
