Applying advanced linear models in the task of predicting student success

The paper presents a comparative analysis of different linear models based on the Moodle data related to the course Programming 2 at the Department of Informatics, University of Rijeka. The task is to predict student final course success based on student activity represented by the initial set of features. We experiment with several methods with the aim to reduce the feature set in order to extract the most representative features thus increasing prediction accuracy. The interpretation of the obtained predictive models is given.