Analysis of recommendation rethods for learning objects using learning styles

Abstract

This master thesis presents a research about the utilization of Learning Styles as the primary criteria for the construction of a recommender a system for Learning Objects. The primary needed characteristics for the creation of such system are identified, along with which steps are necessary for the usage of Learning Styles during the recommendation process. It is proposed a utility function for the recommendation process that utilizes Artificial Neural Networks. So as to carry these tests, synthetic data databases containing students based on the Felder & Silverman model and learning objects on the IEEE-LOM model were created. The results show possible inherent limitations on the method of Anitha and Deisy for the classification of learning objects to the Felder & Silverman model and also to the capacity of learning styles to be used as the main and only factor for the recommendation of learning material. From the conclusions it is seen that the addition of more elements beyond the learning styles that are also pertinent to the learning process and a bigger study about the implementation and training of the neural network could lead to the necessary path to suppress and even overcome the aforementioned limitations, which are the usage of the method proposed by Anitha and Deisy for converting IEEE-LOM learning objects to the FSLSM and the exclusive usage of learning styles as criteria of recommendation.