Abstract—Massive Open Online Courses (MOOCs) is a new approach to online learning which provides a platform for learning in highly scalable and flexible manner. Many higher education institutes are developing and delivering a wide range of such courses. MOOCs are gaining popularity, however they are prone to early dropout and low completion rate. Students registering in MOOCs are different than traditional higher education students in terms of age, education background and motivation. These differences pose challenges in understanding their intent in registering for these courses. In order to improve students’ retention in online learning environment, it is necessary to predict the likelihood of dropout. Timely and proper academic intervention could help struggling students during the course. In this paper, we used MOOCs dataset as a case study to predict student dropout based on the count of online activities. We used classification methods that have been utilized in the field of education domain and are suitable for imbalanced dataset. The machine learning algorithms used in our experiments are: Naive Bayes, Random Forest, Logistic Regression and K Nearest Neighbor. Our results show that techniques used in this study are able to make predictions of dropout, and Logistic Regression outperformed other classifiers with maximum accuracy.
Index Terms—Learning analytics, MOOCs, machine learning, data mining, prediction.
Rahila Umer, Teo Susnjak, and Anuradha Mathrani are with Institute of Natural and Mathematical Science, Massey University, Auckland, New Zealand (e-mail: r.umer@massey.ac.nz, T.Susnjak@massey.ac.nz, A.S.Mathrani@massey.ac.nz).
Suriadi Suriadi is with Queensland University of Technology, Brisbane, Australia (e-mail: s.suriadi@qut.edu.au4).
[PDF]
Cite: Rahila Umer, Teo Susnjak, Anuradha Mathrani, and Suriadi Suriadi, "Prediction of Students’ Dropout in MOOC Environment," International Journal of Knowledge Engineering vol. 3, no. 2, pp. 43-47, 2017.