<![CDATA[Identifying students’ learning styles is essential for developing adaptive educational systems that enable personalized learning experiences. Traditional methods, however, often rely on static questionnaires and lack adaptive mechanisms that can learn from data. This study proposes a reinforcement learning framework for automated learning style identification using the SARSA algorithm, grounded in the Felder–Silverman Learning Style Model (FSLSM). The SARSA agent learns behavioral patterns from students’ responses to the Index of Learning Styles (ILS) questionnaire. Data were collected from 799 university students across four FSLSM dimensions: active–reflective, sensing–intuitive, visual–verbal, and sequential–global. To address class imbalance, class weighting was applied during training to ensure balanced representation across categories. The model achieved strong performance across all dimensions, with accuracy ranging from 0.92 to 0.98, and precision, recall, and F1-scores consistently above 0.90 across cross-validation folds (k = 3, 5, 10). These results confirm the model’s reliability and generalization capability. Validation on a separate dataset of 45 students yielded a Cohen’s Kappa coefficient of 1.0, indicating perfect agreement with the FSLSM-based ground truth. Overall, the findings highlight the effectiveness of the SARSA algorithm in accurately identifying learning-style patterns and demonstrate its potential for integration into adaptive and personalized learning systems.]]>
Copyrights © 2026
