Abstract:

In order to estimate future academic performance or to detect at-risk students in blended learning, we proposed a methodology to assess transition of relative academic performance compared to other learners.

Learning analytics utilizing learning histories from LMSs and similar systems are generally considered effective for tracking learners' progress and identifying at-risk learners. However, compared to full online courses where almost all learning activities and their outcomes can be tracked, blended learning environments, where only partial learning activities can be captured, fail to provide sufficient information to accurately assess learners' progress. This makes it difficult to track academic performance trend of individual students, predict future performance based on the trend, or detect students who may become at risk in the future.

On the other hand, blended learning has the advantage of using common materials and following a same schedule, making it easy to compare one's current learning progress and academic performance with other learners. Therefore, this study proposed a methodology to access transition of relative academic performance compared to other learners. We also conducted a preliminary investigation to confirm the feasibility of the proposed methodology using learning history of blended language courses which consist of e-learning and face-to-face class as an example.

The targeted blended language course in this study was a Chinese language course for freshmen as a second foreign language in a university, and conducted through 2 semesters. Students learned new contents with on-demand videos and practiced them with quizzes provided via a mobile application, and then performed advanced practices in a face-to-face class every week. This time, we assumed the number of attempts required to answer the quiz correctly as absolute academic performance of a student, calculated the relative academic performance based on the difference from the average performance of all students for every period, and confirmed its transition as a comparative transition of each student.

We applied the proposed methodology to several students and confirmed that while some students maintained stable performance, others gradually declined in performance through a semester. This means that the proposed methodology can predict future academic performance based on past trends, and if students showing a significant decline in academic performance are identified, they can be detected as at-risk students.

We are planning to examine patterns in academic performance trends, and differences in trends across learning content, and analyze predictions of learners' strengths and weaknesses, and also investigate the relationship between learning behaviors and academic performance.

Keywords:

Learning Analytics, Blended Learning, Academic Performance, Comparative Transition.