Inactive Behavior Analytics In On-Site Lectures

Detection of at-risk students is a fundamental issue in enhancing learning supports, and has been proposed based on students' learning activity in learning analytics. However, it is not clear which activity we should focus on to detect at-risk students such as low performance students. In this study, we proposed a clustering-based method for at-risk student detection based on three main clusters of students: inactive, passive, active students. Our method focused on reading behaviors and action behaviors in an e-book system. In addition, we consider which period of learning activities is effective for detecting at-risk students. The learning logs of 289 students of Cyber-Security course were collected for our analysis. In our comparison at different moment during the lecture, we found that the cluster of inactive students detected after 35 minutes of lecture got significant lower grades than other students, when the lecture was not too short nor too easy.