Predicting Learning Styles Using Machine Learning Classifiers

Intelligent Tutoring defines the use of intelligent systems to personalize the tutoring and mimic the behavior of human tutors to improve the learning environment and enable learners to study and discuss topics in natural language, to have a deeper understanding of the topic. Various models are proposed to achieve intelligent tutoring; these models aim to predict the suitable teaching strategy and best learning style for each learner. However, these models have not covered all the student’s behaviors and preferences. Therefore, more analysis is needed to understand learners’ needs and examine the lack of existing systems to provide more efficient intelligent systems for effective one-to-one personalized learning. This research proposes a new model based on students’ behaviors and skills. A dataset of e-learning student reactions is used, which is a large-scale dataset for posts and reactions created by students on the e-learning platform. Three different classification methods namely Classifier Chains, Binary Relevance, and Label Powerset are applied to make a model for learning styles prediction and provide the best experience for each learner. Label Powerset achieves the best results of F1 Score (0.93) among all the binary classifiers that used the problem transformation method.