Predictors of mental health literacy among adolescents using a random forest model: a machine learning approach

Abstract

Mental health literacy (MHL) plays a critical role in early identification and help-seeking for mental disorders, especially among adolescents, a group facing a rising burden of depression and psychological distress globally. Traditional statistical approaches may overlook complex interactions among predictors. This study aimed to identify key predictors of MHL using a machine learning-based approach, specifically the Random Forest Model (RFM). A cross-sectional dataset comprising 1707 Malaysian adolescents with complete responses on 10 sociodemographic and psychosocial variables was analysed. MHL was dichotomised into Adequate and Inadequate categories. The dataset was split into training (80%) and testing (20%) subsets. Model development, hyperparameter tuning, and performance evaluation were conducted using R statistical software, with 5-fold cross-validation repeated three times. The final model was assessed on the test dataset using accuracy, sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), F1 score, and Area Under the Curve (AUC). The final RFM, configured with 100 trees (mtry = 4, min.node.size = 1), achieved an AUC of 0.943 on the test dataset. The model achieved an accuracy of 87.1%, sensitivity of 90.0%, specificity of 87.0%, PPV of 17.0%, and F1 score of 0.29. Age emerged as the most important predictor, followed by Parental supervision and alcohol consumption. The RFM demonstrated high overall discrimination in classifying MHL levels. However, the relatively low precision suggests limited accuracy in identifying true positive cases, likely due to class imbalance, and findings should therefore be interpreted cautiously.