Investigating Students’ e-Book Reading Patterns with Markov Chains

Abstract

In this paper, we analyze students’ e-book reading patterns by using MarkovChains (MCs). We used click-stream data of 236 students while they read 7 differentcontents shared by the instructor across different weeks of the course. To analyze readingpatterns, we first clustered students independently based on their interaction with eachcontent. We grouped students in None, Low, Medium, and High clusters. Then by usingMCs, we calculated cluster transition probabilities between different contents. We alsovisualized these patterns and applied a prediction algorithm to predict students’ readingpatterns. Results revealed that students are likely to follow the same reading patterns acrossthe semester. In other words, if a student reads less in the first content, s/he is likely to readless during the rest of the semester. We also found that transition data could be used topredict students’ further reading behaviors. The developed model can be used to plan anintervention system for at-risk students. Visualization of these transitions may help a teacherto understand how well students use contents.