A hybrid unsupervised learning framework for Big Five personality-based profiling and mental health risk inference among undergraduate students
DOI:
https://doi.org/10.18488/76.v13i4.5373Keywords:
Big five personality traits, Educational data mining, Hybrid clustering, Mental health risk inference, Psychological profiling, Unsupervised learning.Abstract
This study describes the development of a hybrid unsupervised learning approach for Big Five personality profiling and the inference of mental health risk among college students. The goals were to identify meaningful patterns of personality clustering, detect atypical profiles, and relate these patterns to non-clinical measures of stress, anxiety, depression, and psychological well-being in academic settings. Our final design combined DBSCAN for outlier detection and K-means for stable clustering, followed by principal component analysis for visualization. This approach was applied to a real dataset of 1,214 student responses obtained with a well-validated, 50-item instrument measuring the dimensions of Extraversion, Neuroticism, and Agreeableness. The findings showed that the optimal clustering configuration was obtained at K = 2, with a silhouette score of 0.304, yielding three interpretable groups: an atypical/outlier cluster, an emotionally vulnerable cluster, and an adaptive/balanced cluster. High scores on measures of Neuroticism were strongly predictive of negative mental health risk patterns. In contrast, high scores on measures of Extraversion, Agreeableness, and Conscientiousness were associated with high levels of psychological well-being. Statistical validation using Spearman correlation and one-way ANOVA confirmed substantial differences among clusters, especially in measures of anxiety and psychological well-being. Overall, these results support the utility of hybrid unsupervised learning methods for student profiling, early detection of vulnerability patterns, and the design of individualized counseling and academic support programs within non-clinical college settings.
