Uma abordagem híbrida baseada em redes bayesianas e ontologias para modelagem do estudante em sistemas adaptativos e inteligentes para Educação

Abstract

New technologies emerge everyday with the aim of supporting the development of tools that are more adaptable to the particular needs of each user. When these technologies are associated with the educational environment, such adaptations allow for the improvement of the teaching process. One of the computational resources that strives to improve these processes are Adaptive and Intelligent Educational Systems (AIESs), which possess the capacity to alter its characteristics in order to attend to the users’ needs. These systems use Student Models that are capable of monitoring important student characteristics and make appropriate adjustments, which are directed toward supporting and improving learning. Studies available in the literature propose models capable of evaluating the student’s performance. However, important questions such as knowledge and behavior are rarely analyzed together. In this manner, this study presents the definition of a hybrid approach for student modeling, which is capable of identifying their performance on an AIES. The proposed approach is based on information concerning the level of knowledge and behavior of the students for defining their real cognitive state. In order to realize this task, the approach combines the use of Bayesian Networks and ontologies, with the aim of creating a model that is dynamic, probabilistic, independent of domain, extensible and reusable. It is also proposed an extension to make the model an Open Student Model, which allows for the visualization of the main capacities and limitations of those involved in the learning process. In addition, a case study was performed by means of the implementation and integration of the proposal into a real teaching environment. Under the intent of evaluating the efficiency of the model, four experiments were carried out with students enrolled in face-to-face courses, to verify the correlation between the values inferred by the model and the student's knowledge. Behavioral issues were also verified by means of statistical tests on the data obtained from the experiments. Results showed that the proposed model is sufficient to determining a level of performance that corresponds to the skills and limitations of the students. Therefore, the proposed approach proved to be effective and adequate for modeling the students in AIESs.