Predicting Emotional Well-Being of Undergraduate Students Using Machine Learning Approaches

This paper investigates the emotional well-being status of 719 undergraduate students through 10 questionnaires co-relate with psychology and comprising negatively phrased items. Emotional well-being plays an important role in academic activities, social connect, and long-term mental well-being. The status of emotional well-being among various undergraduate students is a critical determinant of academic activities, personal development, and long-term state of mental health. The present paper applies machine learning methods to predict students' emotional well-being using questionnaire dataset based on a Likert scale. Since all items were negatively worded, responses were reverse-scored to compute overall emotional well-being indices. Both regression and classification models were implemented to predict continuous well-being scores and categorize students into low, medium, and high emotional wellbeing groups. Additionally, unsupervised learning approaches such as K-Means, Hierarchical Clustering, and DBSCAN were applied to discover hidden patterns in student responses. Results indicate that Random Forest Regression achieved the highest predictive accuracy ($\mathbf{R}^{\mathbf{2}} \boldsymbol{=} \mathbf{1}$), while Logistic Regression provided robust classification (Accuracy = 0.99). Clustering analysis revealed the presence of compact groups of students sharing similar well-being profiles. These findings suggest that data-driven approaches can provide early insights into students' emotional health, enabling universities to design targeted interventions.