Leveraging Machine Learning Algorithms for Depression Prediction in College Students Using PHQ-9 Scores

Depression among college students significantly affects their well-being and academic performance. Using PHQ-9 responses and Questionnaire, this study uses models of machine learning - including Logistic Regression, XGBoost, Support Vector Machine and Random Forest in detecting depression. Key indicators were extracted to evaluate the models using recall, accuracy, precision, and F1-score showing that the Random Forest model attained the highest performance with an accuracy of 98.00% and an F1-score of 98.68%. XGBoost closely follows with 97.00% accuracy and 97.00% F1-score, while SVM also performed well at 96.59% accuracy. In contrast, Logistic Regression lagged significantly at 80.68% accuracy. These findings illuminate the effectiveness of advanced machine learning models in the early detection of depression among college students, suggesting their potential integration into campus mental health service.