Abstract

Learning style refers to the term describing the attitudes and behaviors that define the way people like to learn. Determining the learner’s most suitable learning style can promote motivation and academic achievement. As educational data is increasingly accessible, machine learning methods present potential solutions to automatically discover learning preferences and encourage adaptive instruction approaches. This research is an application of a machine learning-based method to predict the learning styles in the VARK model. Two supervised machine learning models (artificial neural networks and random trees) were developed and tested based on a dataset obtained with a sample of students in the Gifted Unit of Northern Border University. The effectiveness of the presented machine learning models was evaluated by several quantitative measures, such as the accuracy, kappa statistics, and ROC-AUS analysis. The findings reveal that the models are successful in capturing the nonlinear relationships among learner attributes and the categories of learning styles. The random tree algorithm was found to be superior in most evaluation metrics. Such results confirm the technical capability and strength of machine learning methods to provide automated learning-style prediction. Also, the research offers a comparative empirical analysis that justifies the incorporation of intelligent classification models in data-driven learning systems, and offers a baseline to further studies about scalability, feature enhancement, and adaptive learning systems.

Statistics