ThomaSIGHTS: Evaluating Nursery Students’ Learning Styles through Body Language and Engagement

Identifying diverse learning styles in early childhood education is a significant challenge that can hinder academic progress. To address this, this study presents ThomaSIGHTS, a system developed to analyze nursery students’ body language and engagement for effective learning style prediction. The system utilizes YOLO and DeepSORT algorithms for body language detection, the Random Forest algorithm to predict the percentages of visual, auditory, and kinesthetic learning styles, and Hybrid Filtering to suggest tailored teaching approaches. Evaluations conducted through a survey involving IT professionals and nursery teachers, adhering to ISO/IEC 25010 standards, resulted in high scores for Functional Suitability (4.17), Usability (4.35), Reliability (4.33), and Performance Efficiency (4.36). Additionally, the system achieved an mean precision of 82.08% an average recall of 59.42%, an average F1 score of 68.25%, an overall accuracy of 62.17%, and an average precision (AP) of 79.5%. The confusion matrix results indicated the identification of 4,305 true positives and 3,195 true negatives, demonstrating strong classification capabilities, although 2,272 false positives and 917 false negatives revealed areas for improvement. These findings demonstrate that ThomaSIGHTS is a reliable tool for detecting students’ body language, predicting learning styles percentages, and suggesting tailored teaching approaches, thereby enhancing classroom engagement and addressing the educational needs of nursery students.