Time Series Classification for Eye Tracking-Based Visual-Verbal Learning Style

Classifying learning styles is crucial for personalized education. Eye tracking has been used successfully to classify visual-verbal learning style. However, previous studies on classification of visual-verbal learning style still relied on proprietary eye movement analysis software. Thus, limiting feature extraction from eye movement data. To address this research gap, we investigated the use of raw time series eye tracking data and machine learning to classify visual-verbal learning style using the Felder-Silverman Learning Style Model (FSLSM). This study evaluated several machine learning methods to classify visual-verbal learning styles based on eye tracking data. We also systematically evaluated the effect of various window sizes on the classification accuracy of machine learning models. Experimental results indicated that the Supervised Time Series Forest (STSF) as the best performing classifier with an accuracy of 97.70% (Z = −2.812, p = 0.005). We also found a window size of 10 demonstrated superior performance compared to other window sizes. Our findings demonstrate that features extracted from raw eye tracking data significantly enhance classification accuracy, highlighting the potential of time series eye tracking data for classifying visual-verbal learning styles. This study serves as an initial step toward enhancing data-driven personalized education and adaptive learning technologies.