온라인 강좌의 LMS 로그 데이터를 활용한 기계학습 기반 학업성취도 예측 변수 탐색 = Exploring predictors for academic achievement using machine learning on LMS log data from the online course

다국어 초록 (Multilingual Abstract)

This study aimed to identify online learning behavior patterns and explore key variables for predicting academic achievement by analyzing LMS log data. Data were collected from 48 students who enrolled in a 15-week online teacher education course at A Graduate School of Education. From 18,270 rows of log data, a total of 123 features—capturing both weekly learning activities and overall course trends—were used to build the prediction model. Academic achievement prediction was conducted using the machine learning technique elastic net, and the findings were twofold. First, an analysis of weekly learning patterns revealed that behaviors differed according to the type of learning content (e.g., video, link, file).
Furthermore, contrary to concerns about cramming or procrastination, most learners did not engage in cramming or procrastination, with approximately 15% of learners exhibiting cramming behavior. Second, the predictive modeling yielded 19 variables that were reliably selected for the final model. Key predictors of higher academic achievement included: a higher frequency of content viewing and material downloads in specific weeks; an earlier start to weekly learning; a greater number of completed assignments; more frequent content access during the final-exam preparation period; and enrollment in an A-F grading system. Based on these results, the study discusses the significance of the selected variables within the course context, as well as the advantages and implications of utilizing machine learning techniques in learning analytics.