Automatic detection methods of students’ learning styles in learning management system / Li Ling Xiao

• Auteur principal: Li , Ling Xiao
• Format: Thèse
• Publié: 2019
• Sujets: QA75 Electronic computers. Computer science

Online learning has become a common phenomenon nowadays. Many distance-learning systems or platform distribute educational resources online. Meanwhile, in order to satisfy students’ learning experience and to improve learning effectiveness, students’ characteristics should be considered, from the point of view of knowledge level, goals, motivation, individual differences and many more. The focus of this thesis is on the learning style as the criterion. Students are characterized according to their own distinct learning styles. Identifying students’ learning style is vital in an educational system in order to provide adaptivity. The first step towards providing adaptivity is knowing students’ learning style. Past researches have proposed various approaches to detect the students’ learning styles. However, the results obtained from the past researches have been disparate in terms of precision. Broadly speaking, the existing automatic detection approaches are only able to provide satisfactory results for specific learning style models and/or dimensions, or even only work for certain educational systems. The aim of this thesis is to study on an automatic detection of learning styles to address the existing issues, mainly focusing on improving the precision of detection. The first proposed approach for automatic detection is the construction of a mathematical model from the analysis of students’ learning behaviour. This approach specifically explores the relationship between students’ learning behaviour and their learning styles. However, the precision of the results obtained from this approach show only moderate precision, equivalent to the results obtained from the past researches. A possible reason for this is that the approach is designed for general applicable model with relatively loose conditions. To further improve the precision of the detection, this thesis next proposes tree augmented naïve Bayesian network for automatic detection of learning styles. Bayesian network has emerged as widely a used method in this field but, then again, tree augmented naïve Bayesian network has the ability to improve the classification precision. The performance of tree augmented naïve Bayesian was evaluated in an online learning environment called Moodle. The experimental results are very encouraging. The proposed tree augmented naïve Bayesian network method is able to provide good results for all dimensions of Felder-Silverman learning style model, which can be seen as an appropriate method to detect learning styles with higher precision.