Uma abordagem Multiobjetivo para a Recomendação de Objetos de Aprendizagem no contexto Educacional.

Currently, one may find large repositories on the Internet that can be used as a source of resources for e-learning. In order to facilitate the use of these contents, they are organized as Learning Objects (LOs), which, in addition to the content itself, they hold several metadata to describe them. Due to the large volume of materials available, it becomes difficult to choose those LOs that best meet the needs of each individual. To suggest the most suitable LOs, we rely on Recommendation Systems, that are responsible to find and recommend the most useful items to students. In this context, the present work proposes a multiobjective optimization approach for the LOs recommendation problem. In this problem, there is a set of concepts to be taught, according to an Instructional Design, a repository of LOs, in which each LO addresses one or more concepts, and also pedagogical dependencies, that enforce certain concepts to be presented in advance to others. In this problem, we do not seek for solutions that necessarily cover all established concepts. In fact, we aim to obtain solutions that fulfill two distinct objectives, namely: i) lower cost, given by the sum of the cost of each LO present in the solution; ii) greater number of concepts covered, respecting the established pedagogical dependencies. To solve the problem, we used the NSGA-II, a multiobjective Genetic Algorithm implemented in the MOEA framework. The problem instances were artificially generated, simulating repositories of 1000 LOs, courses with 200 concepts, with up to 20 dependencies. The results obtained covered sets of different configurations of solutions, with different numbers of concepts covered and costs of the solutions, which brings greater possibilities to deliver content in accordance with the objectives, needs and reality of the students. [ABSTRACT FROM AUTHOR]