Mining activity grades to model students' performance

E-learning systems have major benefits but also pose major challenges. One of these is how to do a good job of tutoring students without face-to-face contact. This calls for the interpretation of large quantities of data generated as a result of the activities performed by students, which e-learning platforms collect and store. These data are also potentially very useful for preventing student dropout. We propose the use of knowledge discovery techniques to analyse historical student course grade data in order to be able to predict in real time whether or not a student will drop out of a course in the future. Logistic regression models are used for the purpose of classification. Experiments conducted with data on over 100 students for several real distance learning courses confirm the predictive power of our proposal that outperforms other existing approaches in terms of accuracy.