Machine Learning-Based Prediction of University Students’ Mental Health: Integrating Developmental and Psychosocial Factors in Higher Education Analytics

Due to their impact on psychological well-being, educational engagement, and long-term educational outcomes, mental health problems among university students have received increasing attention. The goal of this research was to use machine learning algorithms to predict mental health conditions among university students and to assess how developmental, demographic, academic, and psychosocial factors affect those predictions. A publicly available dataset containing 101 university students was used to perform analyses using logistic regression, random forests, and gradient boosting algorithms, with the following variables included: age, gender, marital status, cumulative grade point average (CGPA), anxiety, panic attacks, and treatment-seeking behavior. The data were preprocessed, features were engineered, and an 80/20 train-test split was used to assess the predictive performance of the models. The results showed that gradient boosting achieved the highest accuracy (88%) in classifying targets, followed by random forest (84%) and logistic regression (76%), and the confusion matrix analyses further confirmed that the predictive models were stable, as evidenced by low false-positive and false-negative classifications. The most important predictors were anxiety and marital status; CGPA was a relatively weak predictor. Ensemble ML methods identified complex relationships in student mental health, and this study shows how to use these methods to provide an exploratory, predictive framework for early identification and screening of mental health issues, and to support strategies for students through data-informed decisions in universities.