Acquisition of learning style of MOOC learners from large-scale learning log-data

Abstract

In this paper, we consider a method for acquiring learning style of MOOC learners. Learning style is a pattern on how learners learn. It is useful for helping the learners in both not to dropout from the course and improving the learning materials. When a learner learns on a MOOC system, the log-data are recorded on the system, such as the learner's operations and system's responses. We focus on transition events that occur when the learner moves from a page to another page. MOOC learning materials consist of two types of pages, that is video-pages and problem-pages. Each of transition events is described by ``transition-feature'' which consists of a 3-tuple; ``quantity'', ``from page-type'' and ``to page-type''. We clustered the MOOC learners with transition-feature vector and compared the generated clusters based on the pass rate of the course. In the process we extracted the clusters of learners which a previous research had been suggested to be exist. The results of our analyses indicated: 1) there was a correlation between the pass rate and the number of transition from a video-page to the next video-page. 2) There also was a correlation between the pass rate and the number of transition from a video-page to the previous video-page, then we obtained the hypothesis that the learners who watch video-pages more were less likely to dropout.