Early-Stage Identification of Depression in Traditional Students Age Group Using Machine Learning with Composite Parameters

Student mental health is an increasing issue in schools around the globe, and depression is one of the more common conditions affecting academic performance and overall well-being. In this study, we mainly focuses on a machine learning approach for predicting depression among students by using demographic, academic and lifestyle features. For this, we have done data analysis of a dataset of student records containing 17 features including age, gender, CGPA, sleep duration and several mental health related indicators. We tested six machine learning algorithms in total-Logistic Regression, Random Forest, Decision Tree, Support Vector Classifier (SVC), XGBoost, and AdaBoost. After hyperparameter tuning, Adaboost performed the best, achieving an accuracy of 84.80%, an F1-score of 84.71% and ROC-AUC of 92.54%. These results were slightly better than what has been reposted in similar recent studies. From the feature importance analysis factors such as sleep duration, academic pressure, and family mental illness history were the strongest predictors. This research will assist in making early intervention strategies for student mental health support systems in educational institutions.