Abstract: In MOOCs, the only medium for student communication is the online platform and associated forums. Students' interactions provide researchers with insights into their online behaviors and learning habits. Prior studies have shown that these online behaviors reveal student learning processes and also can predict MOOC student dropout and completion performance. However, existing studies focus primarily on either text data from forum posts or log activity (clickstream) data. % Most have not taken full advantage of the multiple types of data that may be available including the semantic content from the discussion forum posts.

In this work, we explore students' online platform behaviors based upon an educational data mining MOOC. We first examine how content from student posts relates to their final performance. Then we analyze their online transitions between problem-solving, watching lectures, and the different types of comments (e.g. questions, answers, off-topic replies, etc.) they made on the forum. The results indicate that students' post content relates to their performance and we are able to find significant differences between the online action sequences of high and low-performing cohorts. These findings will help MOOC instructors detect what behaviors contribute to student learning, what hampers their success and provide guidance to students.

Short Bio: Yiqiao is a Ph.D. student from the ArgLab. His research area focused on educational data mining, specifically, on combining natural language processing and social network analysis into educational data mining.