KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 6 Serial Number 40 2025 12 10 Predicting Adolescent Emotional Dysregulation Using Ensemble Machine Learning Models Integrating Family, School, and Digital Behavior Indicators Predicting Adolescent Emotional Dysregulation Using Ensemble Machine Learning Models Integrating Family, School, and Digital Behavior Indicators 1 11 10.61838/kman.jayps.4904 EN Journal Article 2025 07 25 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> The objective of this study was to develop and evaluate an ensemble machine learning framework for predicting adolescent emotional dysregulation by integrating family, school, and digital behavior indicators.</p> <p><strong>Methods and Materials: </strong>This study adopted a cross-sectional predictive design and was conducted among secondary school adolescents in Mexico. Data were collected using validated self-report instruments assessing emotional dysregulation, family functioning, school climate, and digital behavior patterns, alongside demographic variables. After data preprocessing, including normalization, imputation of missing values, and feature selection, multiple supervised machine learning models were developed. These included linear regression, support vector regression, random forest, gradient boosting, and a stacked ensemble model combining heterogeneous base learners. Model training and evaluation were performed using repeated k-fold cross-validation to ensure robustness and to minimize overfitting. Predictive performance was assessed using root mean square error, mean absolute error, and explained variance.</p> <p><strong>Findings: </strong>Inferential analyses demonstrated that ensemble-based machine learning models significantly outperformed traditional linear and single-algorithm approaches in predicting emotional dysregulation. The stacked ensemble model achieved the highest explained variance and the lowest prediction error. Digital behavior indicators accounted for the largest proportion of predictive importance, followed closely by family-related factors and school-related variables. At the individual predictor level, problematic digital use, family conflict, parental warmth, nighttime device use, and teacher support emerged as the most influential features. Risk-related predictors showed positive associations with emotional dysregulation, whereas relational and contextual support variables showed negative associations.</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/jayps/article/download/4904/9304