Using Social Network Analysis Metrics of Virtual Forums to Predict Performance in e-Learning Courses

Abstract

The present article proposes the use of social network metrics extracted from forums interactions in distance education courses in order to predict students failing. Eight centrality metrics from forums were used as input information for training and testing five different classifiers able to early predict at-risk students. The initial findings indicate these attributes are informative and useful for prediction, however predictive models performance vary considerably across courses and depending on the amount of data collected.