AI-Enabled Machine Learning Framework for Depression Risk Prediction and Mental Health Trend Analysis in the United States

Abstract

Mental health disorders such as anxiety and depression are rapidly increasing and represent a major challenge for modern healthcare systems. This study proposes a machine learning based framework for early detection and analysis of depression risk using behavioral and socio-demographic indicators. A dataset containing features such as age, sleep hours, stress level, income, and mental health days was used to train and evaluate multiple machine learning models, including Logistic Regression, Random Forest, and XGBoost. The results show that ensemble learning approaches outperform traditional models, with XGBoost achieving the highest predictive accuracy. Model evaluation using confusion matrix, ROC curve, and precision recall analysis demonstrates strong classification performance. Feature importance and explainable AI analysis using SHAP reveal that stress level and sleep hours are the most influential predictors of depression risk. Trend analysis across age groups and state-level risk visualization further highlights demographic and regional variations in mental health patterns. The findings demonstrate the potential of machine learning for large-scale mental health surveillance and data-driven public health decision-making.