Ensemble Machine Learning Approach for Identifying Determinants of Student Satisfaction

This research focus on a comprehensive data-driven approach to analyzing student satisfaction using both unsupervised and supervised machine learning techniques. A dataset of 1,000 university students, including 38 survey-based features across academic experience, infrastructure, and career services, was utilized. Principal Component Analysis (PCA) was employed for dimensionality reduction. Clustering techniques including K-Means, DBSCAN, and Hierarchical Clustering identified distinct satisfaction profiles. K-Means (k=3) delivered the most interpretable structure and was selected for subsequent cluster profiling. A Random Forest classifier trained on normalized features achieved a 96% prediction accuracy, with F1-scores ranging from 0.94 to 0.97. The study culminates in targeted recommendations for institutional strategy based on cluster characteristics, illustrating the utility of ensemble learning in educational analytics.