Machine learning-driven prediction of college students’ mental well-being: Insights from social media usage

This study examines the relationship between social media usage and mental well-being among college students in Tamil Nadu, employing a descriptive and diagnostic research design with stratified random sampling. Data were collected from 384 students using the Warwick-Edinburgh Mental Well-Being Scale and social media usage metrics. Findings reveal that students spend an average of 4 h daily on social media and use mobile applications approximately 44 times per day. Notably, 31.8% of students reported very low well-being and 15.4% reported below-average well-being, indicating that nearly half the sample is at risk. Non-parametric analyses identified significant associations between app categories, gender, and income with well-being. Machine learning models were applied to predict student well-being, with the Support Vector Classifier achieving the highest accuracy (76.18%) and balanced F1-score. A predictive dashboard was developed to support early intervention and counseling. This study is among the pioneer to apply AI-driven predictive modeling to the mental well-being of Tamil Nadu college students, integrating social media usage data with well-being measures. By combining traditional statistical analysis with machine learning dashboards, it provides a unique evidence base for region-specific interventions and policy recommendations aligned with NEP 2020.