A dropout predictor system in moocs based on neural networks

Abstract

Massive open online courses, MOOCs, are a recent phenomenon that has achieved a tremendous media attention in the online education world. Certainly, the MOOCs have brought interest among the learners (given the number of enrolled learners in these courses). Nevertheless, the rate of dropout in MOOCs is very important. Indeed, a limited number of the enrolled learners complete their courses. The high dropout rate in MOOCs is perceived by the educator’s community as one of the most important problems. It’s related to diverse aspects, such as the motivation of the learners, their expectations and the lack of social interactions. However, to solve this problem, it is necessary to predict the likelihood of dropout in order to propose an appropriate intervention for learners at-risk of dropping out their courses. In this paper, we present a dropout predictor model based on a neural network algorithm and sentiment analysis feature that used the clickstream log and forum post data. Our model achieved an average AUC (Area under the curve) as high as 90% and the model with the feature of the learner’s sentiments analysis attained average increase in AUC of 0.5%.