A Stacked Ensemble Framework through Synergistic Modeling for Adolescent Delinquency Classification

Abstract

This study emphasizes how ensemble approaches can improve the precision and resilience of predictive models in assessing social problems like adolescent delinquency. To classify adolescent delinquency, this study looks into how effective a stacked ensemble framework is. The performance of the stacked ensemble model is compared to individual classifiers using data on different risk factors linked to delinquent behavior. The evaluation metrics that are used are AUC, F-score, and the area under the receiver operating characteristic curve (AUC). The stacked ensemble framework constantly outperforms individual classifiers, as evidenced by the high AUC scores it achieves, which range from 0.97 to 1.00 across all classes of delinquency severity. Moreover, in terms of classification accuracy and F-score, the proposed ensemble model outperforms individual classifiers, demonstrating greater discriminative power. Additionally, the stacked ensemble exhibits improved generalization and robustness, proving its effectiveness in locating underlying patterns in the data and lowering the chance of overfitting. The classification of adolescent delinquency is significantly impacted by the ensemble framework's superior performance, which implies that ensemble techniques like stacking can improve predictability and accuracy in identifying adolescents who are at risk at an early stage. The results demonstrate the potential of the suggested stacked ensemble approach to enhance predictive performance and provide reliable classification across a range of delinquency severity levels, thereby validating its efficacy in classifying juvenile delinquency.