Abstract

Artificial intelligence holds significant potential for enhancing adaptive learning environments. However, effective personalization requires a deep understanding of individual learner characteristics, particularly their preferred learning styles. This study presents an Artificial Neural Network (ANN) - based model, aligned with the VARK framework (Visual, Auditory, Reading/Writing, Kinesthetic), to identify student learning preferences using survey data collected from 700 students across schools, colleges, and universities in Bangladesh. A hybrid architecture combining multi-label classification and multi-output regression was employed to predict both the dominant learning styles and the degree of preference for each. The ANN outperformed traditional machine learning algorithms - including Support Vector Machine, Random Forest, Decision Tree, and K-Nearest Neighbors - achieving an F1-score of 0.92 and R2 score of 0.96. Performance further improved with the integration of K-Means clustering, boosting the F1-score to 0.96. The regression component of the model provides a percentage-based prediction of how strongly a student prefers each learning style, offering a more granular and nuanced understanding of individual preferences. Compared to conventional approaches, this multiheaded approach is more flexible and informative, enabling the early identification of learning styles and facilitating the development of personalized educational content prior to course delivery.

Download

Share

COinS