Comparative Assessment of Machine Learning Models for Detection of Mental Health of Students: A Case Study of Undergraduate Students

Early identification of anxiety and depression in college students is essential for timely intervention, yet conventional screening methods often rely on subjective reporting and consume considerable resources. This study administered a 20-item Likert-scale questionnaire to 142 undergraduates from a rural engineering institution in Maharashtra, India, covering aspects such as sleep quality, social engagement, and academic stress. The data were normalised and split into training (80%) and testing (20%) sets. Six classifiers—Logistic Regression, Support Vector Machine (SVM), Naïve Bayes, Decision Tree, Random Forest, and K-Nearest Neighbours—were trained and evaluated using accuracy, precision, recall, and F1-score metrics. Logistic Regression and SVM achieved the highest performance with 96.6% accuracy and an F1-score of 0.964, followed by Naïve Bayes (93.1%), Decision Tree (86.2%), Random Forest (82.8%), and K-Nearest Neighbours (79.3%). These results demonstrate that interpretable algorithms like Logistic Regression and SVM can reliably screen for mental health risks using self-reported data, offering a scalable, cost-effective approach for campus health services in resource-limited rural settings. The significance of this work lies in addressing the growing mental health crisis among students, where untreated issues contribute to substantial economic losses projected at $16 trillion globally by 2030, by enabling early detection and reducing barriers like stigma and access to care.