Predicting Learning Outcomes with MOOCs Clickstreams

Chen-Hsiang Yu1, Jungpin Wu2, Aa-Chi Liu1
1Department of Information Engineering and Computer Science
2Department of Statictics
Feng Chia University
#100 Wenhwa Road, Taichung, Taiwan
Phone: 886-4-2451-7250 Fax: 886-4-2451-6101
Massive Open Online Courses (MOOCs) have gradually become one of the dominant trends in education. Since 2014, the Ministry of Education in Taiwan has been promoting MOOCs programs with successful results. Due to its self-paced mode, however, the low completion rate of MOOCs has recently become the focus of attention. The mechanism to effectively improve the course completion rate continues to be of great interest to both teachers and researchers.
In this study, we generated a sequence of learning behaviors of learners by using their video clickstream records on the MOOCs platform to find patterns in the learners’ cognitive participation. Then, we built practical machine learning models using K-Nearest Neighbor, Support Vector Machine, and Artificial Neural Network algorithms to predict learning performance through student learning behavior. Using these models, we were able to determine the relevance of video viewing behavior to learning outcomes in order to assist teachers in helping learners who need additional support to pass the course.

MOOCs, Clickstream, Behavior Pattern, Machine Learning





Published in: 

2nd Eurasian Conference on Educational Innovation 2019

Full paper download: