Abstract:

Personalised learning systems provide a unique, specific learning path for particular student or a group of students. They can adapt according to learner’s requirements and preferences. They apply traditional information technologies, systems and tools in such a manner which provides learning based on student’s strengths, weaknesses, psychological portrait, pace of learning, learner’s needs and pedagogical methods best suited. Learning content personalisation, learning content type, representation of learning content, content navigation pattern are the main aspects to consider when personalising virtual learning environments. As personalisation is done by personal traits of a learner and by other information related to particular learner, user profiles and user models are used for modelling and storing such kind of information.

In this paper, first, a systematic review of literature on user modelling is done, focusing on static and dynamic user’s learning style models. Then Bayesian approach to learning style modelling is introduced. In first subsection philosophical approach to representation of causality and belief is described – Bayesian models are based on such approach. Then rules of probability theory applicable to Bayesian models are presented. The following subsection is aimed at description of dynamic learning style modelling using probabilistic Bayesian network.

Bayesian network uses data about learner’s past behaviour in web based learning environment for prediction on properties to be used for future personalisation. As a lot of factors extracted from learner’s past behaviour in adaptive hypermedia learning systems determine his/her learning style, review of literature about patterns of learners’ behaviour together with analysis of practical application of behavioural patterns for students learning style identification was done, trying to systematise stereotypical features (patterns) of learners’ behaviour that can be used to conclude a learning style. A list of key factors which probabilistically are related to the particular learning style has been compiled for quick handy use. Simulation of relationships between random key factors for learning style identification using Bayesian probabilistic graphical model is also presented in the paper. Advantages and disadvantages of Bayesian learning style modelling were specified. Finally, conclusions and future trend are presented.

Keywords:

Virtual learning environments, personalisation, dynamic learning styles modelling, Bayesian network, learners’ past behaviour.