Skip to content
Review Open access

Predictive Modelling of Depression among Adolescents and Adults in the United States using Machine Learning

Oct 2026 · International Journal of Modern Science and Research Technology · 0 citations · 17 references

Abstract

Depression is a significant public health problem that has significant personal and community impacts. Because it is a multifactorial and complex phenomenon, and the factors that contribute to its development are interrelated, traditional statistical models may not be able to precisely identify people at risk. Machine learning (ML) is a data-driven method for modelling complex patterns and enhancing depression prediction. Methods The data for this study was obtained from the National Health and Nutrition Examination Survey (NHANES) 2021-2023 to predict depression among adolescents and adults in the United States. The Patient Health Questionnaire-9 (PHQ-9) was used to assess depression, and a score of ≥10 was considered to be depression. Six supervised machine learning algorithms were tested: logistic regression (LR), random forest (RF), naïve Bayes (NB), support vector machine (SVM), eXtreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM). After data preprocessing, feature selection, and imputation, the data was split into an 80:20 ratio for training and testing. Random undersampling of the training data was used to deal with the class imbalance. For model development, five-fold cross-validation and grid-search optimization were used. The accuracy, sensitivity, specificity, precision, AUC and F1-score were used to evaluate the performance. SHapley Additive exPlanations (SHAP) were used to interpret model predictions. All analyses were carried out with statistical software R (version 4.4.0). Results Among the models evaluated, XGBoost showed the best overall predictive performance followed by LightGBM and random forest. SHAP analysis gave an interpretable framework to analyze the contribution of individual predictors to depression predictions, and showed that depression risk is multifactorial. Conclusion Multidimensional population health information offers a promising way to predict depression with machine learning. Ensemble-based algorithms could be especially beneficial for the modeling of complex relationships that underlie depression risk, as XGBoost performed well in this study. Combining predictive modelling with SHAP interpretation can help in the more transparent and informative assessment of depression risk.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.