Estimation of Test Scores based on Questionnaire and Video Viewing Behavior in the Programming MOOC Course

Abstract

The Massive Open Online Courses (MOOCs) offer various types of learners the opportunity to attend university-level lectures. However, since many learners drop out during the learning process, the average MOOC completion rate is usually as low as 10%. For future improvement, MOOCs must grasp the learners' features in the earlier stage and provide appropriate support to each learner. This paper investigates the relationship between learners' characteristics and test scores in the programming MOOC course to recognize different types of learners. Video viewing behavior and the questionnaire information at the beginning of the lecture, i.e., age, programming skill, and keywords in the free description, are analyzed to characterize learners. As the results, it was observed that the repeated learning behavior and later join to the course relates to the higher and lower score, respectively. The information from the questionnaire improves the accuracy of pass/fail estimation before the third week. The characteristic cluster of learners, who could be rescued by offering appropriate support, was also obtained by multiple regression analysis results.