Predicting Student Depression using Machine Learning: A Comparative Study of Logistic Regression and Random Forest

Abstract

Depression is a common mental health concern among university students, typically caused by academic and social demands. The purpose of this study is to investigate the efficacy of machine learning techniques in detecting depression in its early stages. To optimize model parameters, this research used a Kaggle dataset with 502 participants, rigorous data preprocessing, and a 10-fold cross-validation strategy. Two predictive models were created: logistic regression with elastic net regularization and random forest model. The results reveal that the logistic regression model obtained an accuracy of 98%, beating the random forest model’s 92% accuracy. At the same time, feature importance analysis highlighted academic pressure and suicidal ideation as significant predictors. These results highlight data-driven approaches as likely to improve early diagnosis and targeted intervention of mental health issues in universities. Thus, the research informs depression understanding in student communities, providing a comparative perspective to the effectiveness of various predictive models for guiding preventative and support measures.