Applying machine learning techniques to detect student's learning styles

Abstract

Learning styles play a vital role in determining an individual student's best learning methods and suitable types of learning materials. Based on a survey that collects information about learners' learning styles in offline and online environments, chosen from Felder-Silverman Learning Style Model (FSLM) and David Kolb Learning Style Model, we have developed a model to group students based on their characteristics. Specifically, the data collected from 546 learners in 2 universities in Vietnam are put into clustering using the K-Means algorithm, then labelled and classified by Support Vector Machine (SVM) and Extreme Gradient Boosting (XGBoost) algorithm, and finally evaluated based on precision, recall, and F1 score metric to find the number of suitable groups. The results show a new learning model with four different learning styles, each corresponding to a category in the FSLM and David Kolb model, with values on the accuracy, precision, recall, and f1 score equaling 96%. With the strength of combining the theory of all three different learning style models, along with a machine learning model with high accuracy on a rather large data set, the research results promise to make a positive contribution to the problem of personalized learning content, helping learners have the most effective learning experience.