A284: Machine Learning-Based Prediction of Adolescents' Quality of Life

The impact of school, family, and healthy lifestyle behaviors on adolescents' quality of life (QoL) has received increasing scholarly attention. This study aims to develop an interpretable machine learning model for predicting adolescents' QoL based on three key dimensions: school, family, and health-related lifestyle factors. Random Forest, Decision Tree, Gradient Boosted Decision Tree, Ridge Regression, Support Vector Machine, Multilayer Perceptron, AdaBoost, Voting Classifier, and K-Nearest Neighbors were employed to identify the optimal predictive model for QoL outcomes. To elucidate the nonlinear contributions of variables across dimensions, permutation feature importance, partial dependence plots (PDPs), and Shapley Additive Explanations (SHAP) were utilized. A total of 20,237 eligible adolescents (aged 10–25 years) were randomly assigned to training and validation sets for model development and evaluation. The Random Forest model achieved the highest overall performance, with an accuracy of 79.55% on the validation dataset. Within the comprehensive model, the key predictors of QoL included self-regulation (SR), interpersonal relations (IR), aggressive-nervous tendencies (AN), parental co-exercise (PE), sleep duration, and academic skills (AS), with respective contribution weights of 0.0157±0.0039, 0.0142±0.0038, 0.0071±0.0032, 0.0064±0.0031, 0.0050±0.0016, and 0.0043±0.0033. Specifically, improvements in SR, IR, PE, sleep duration, and AS—or reductions in AN—were associated with enhanced QoL among adolescents. Dimension-specific modeling revealed that within the healthy lifestyle dimension, screen time, physical activity, and sleep emerged as the key determinants of QoL. In the family dimension, parental co-exercise, parental relationship quality, and parental education level were identified as the most significant predictors. In the school adaptation dimension, interpersonal relations and self-regulation were found to be the primary influencing factors. This study underscores the effectiveness of machine learning approaches in identifying the multifactorial determinants of adolescents' QoL. The findings highlight that regular participation in physical activity—alongside sufficient sleep, supportive family dynamics, and positive school adaptation—plays a pivotal role in promoting improved QoL outcomes. These results offer empirical evidence to inform educators, policymakers, and public health professionals in the development of targeted interventions aimed at enhancing adolescent QoL.