Mining MOOC Clickstreams: On the Relationship Between Learning Behavior and Performance

Posted date: 

How can we relate the actions that a learner takes during a course (i.e., their behavior) to the amount of knowledge the learner has gained through the course (i.e., their performance)? This fundamental question, pertaining both to education and to learning in general, is the focus of this work. We propose two different frameworks for representing learner behavior – one as a series of events and their associated durations, and another as a series of positions visited while traversing the course content – and apply these frameworks to two of our Massive Open Online Course (MOOC) datasets. In doing so, we are able to extract recurring behavioral patterns (i.e., “motifs”) that are significantly associated with whether a learner has gained knowledge from content or not, and to substantially improve the quality of performance prediction, verifying the efficacy of our methods in relating behavior and performance.

Link to article:  http://arxiv.org/pdf/1503.06489v1.pdf