<![CDATA[The mismatch between teaching methods and individual learning style preferences in e-learning platforms often hinders the effectiveness of information absorption for students. The primary issue lies in the one-size-fits-all learning approach and the inefficiency of identifying learning styles through manual questionnaires, which are subjective and time-consuming. This study aims to evaluate the performance of Random Forest, Decision Tree, and K-Nearest Neighbor (KNN) algorithms in automating VARK (Visual, Auditory, Read/Write, Kinesthetic) learning style classification. The novelty of this research lies in the implementation of the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance in the Read/Write modality, which initially accounted for only 14.2% of the total population. Following the CRISP-DM framework, a balanced dataset of 1,410 records was utilized. Experimental results show that Random Forest and KNN achieved the highest identical accuracy of 97.87%. However, based on stability evaluation through 10-Fold Cross Validation, Random Forest proved to be the most optimal model with the highest Mean CV score of 0.9592, outperforming KNN (0.9503). These findings provide a precise scientific foundation for developing adaptive recommendation systems to deliver personalized and effective instructional materials.]]>
Copyrights © 2026
