Advancing Recidivism Prediction for Male Juvenile Offenders: A Machine Learning Approach Applied to Prisoners in Hunan Province

This study uses machine learning approach to forecast the likelihood of recidivism among male juvenile offenders. The dataset utilized in this study is the Structured Assessment of Violence Risk in Youth (SAVRY) dataset, which was obtained from Hunan Province, China. After conducting a meticulous examination, a variety of machine learning algorithms were evaluated, including Random Forest, Gradient Boosting, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). These models demonstrated a remarkable accuracy rate above 95% when implemented on the test dataset. The implementation of ensemble approaches, hyperparameter optimizers, and feature selection methods resulted in an enhanced level of predictability. Furthermore, Explainable AI was used to assess the fairness and validity of the models. Our results demonstrate that this proposed approach has successfully improved the performance and interpretability of the ML models to predict juvenile recidivism among young offenders.