Mining Learning Management System Data Using Interpretable Neural Networks

The paper presents the work on data analysis of LMS data related to the course Programming 2, one of the introductory courses at the first year of the study of Informatics at the Department of Informatics of the University of Rijeka. In order to improve the course we analyze the data from the Learning Management System (LMS) with the emphasis on some additional activities which are not graded, such as watching video lectures. We are interested whether these activities have positive impact on the student success. The data analysis can objectively evaluate their role, the effect of improvement, and their impact on the learning process. The paper presents discovery of knowledge about the process of learning using batch data analysis performed by Artificial Neural Networks (ANNs). ANNs are not a common method in the field of educational data mining. Although highly accurate, the resulting black-box model is not interpretable, which is a major drawback. For the opening of the ANN black-box model, as well as for other models of this type, a number of agnostic methods have appeared recently, some of which we illustrate in the LMS system analysis.