IMPLEMENTING SUPPORT VECTOR MACHINES IN TEACHING EFFECTIVENESS EVALUATION SYSTEMS FOR DATA-DRIVEN CLASSIFICATION AND ASSESSMENT

Abstract

Student evaluations of teaching generate substantial unstructured qualitative feedback providing nuanced insights into teaching effectiveness, yet these valuable data remain systematically underutilized in higher education institutions due to manual processing limitations and lack of systematic analysis frameworks. This study aimed to develop and validate a Support Vector Machine classification system for analyzing unstructured student comments to identify patterns associated with teaching effectiveness, reduce manual evaluation processing time, and transform qualitative feedback into actionable insights for faculty development. A mixed-methods correlational design examined 4,872 unstructured student comments from the University of Antique Tario-Lim Memorial Campus. Qualitative coding developed a taxonomy aligned with four institutional evaluation dimensions, followed by SVM model training for automated classification. Statistical techniques included feature importance analysis and cross-validation. Research limitations encompassed single-institution scope and dependency on existing evaluation frameworks. The optimized SVM model achieved 87.3% classification accuracy in distinguishing effective from ineffective teaching practices. College of Teacher Education demonstrated highest classification performance at 89.2% among four colleges examined. Feature importance analysis identified instructional clarity references, question responsiveness, and engagement techniques as strongest predictors of teaching effectiveness. Implementation reduced manual evaluation processing time by 68% while maintaining substantial agreement with expert evaluations. Machine learning approaches effectively transform qualitative student feedback into quantifiable insights for teaching effectiveness assessment, demonstrating significant potential for scalable faculty evaluation systems in higher education institutions. Future investigations should examine cross-institutional validation, explore deep learning architectures for enhanced classification accuracy, and investigate longitudinal patterns in teaching effectiveness feedback across multiple academic periods.