A Real-Time Predictive Model for Identifying Course Dropout in Online Higher Education

Abstract

Course dropout is a concern in online higher education, mainly in first-year courses when different factors negatively influence the learners' engagement leading to an unsuccessful outcome or even dropping out from the university. The early identification of such potential at-risk learners is the key to intervening and trying to help them before they decide to drop out. This paper focuses on this challenging problem by providing a predictive dropout model with distinctive characteristics from previous approaches. First, the identification is in real time by providing a daily dropout prediction. Second, a temporal window of variable size is defined to evaluate the likelihood of being a dropout learner at the activity level. Such contributions will serve as a basis for designing and applying intervention mechanisms to reverse the course dropout at-risk situation. The predictive model and the temporal window have been evaluated on data from an authentic online learning setting in two first-year undergraduate courses. We show the accuracy of correctly identifying at-risk learners within activities and the model performance to detect actual course dropout learners.