Domain-Sensitive Learning Analytics Under Adaptive Outcome-Based Education: A Within-Subjects Comparison of Technical and Analytical Courses

Abstract

Learning Management Systems (LMS) are now a vital source of information for evaluating student learning and predicting academic success. Students' involvement can now be evaluated using metrics such as time spent on learning assignments, quiz attempts, and course activities, thanks to the growing use of LMS platforms. Nevertheless, a lot of current research assumes that these engagement metrics have comparable predictive power across academic fields without considering the variations in learning situations among domains. In an adaptive Outcome-Based Education (OBE) setting, this study examines whether LMS engagement indicators show domain-dependent prediction patterns. Sixty undergraduate Business Analytics students who took two courses via Moodle LMS — NoSQL/MongoDB as a technical course and Applied Business Statistics as an analytical course — were the participants of a within-subjects study. As indicators of student achievement, engagement metrics such as activity completion, quiz attempts, and LMS time were investigated. The investigation employed exploratory machine learning classification, Steiger's Z-test, and multiple linear regression. The findings show that the technical course (R 2 = 0.62) has a higher predictive strength than the analytical course (R2 = 0.42). Steiger's Z-test indicated that the variation in prediction strength was statistically significant (Z = 2.12, p = 0.034). Additionally, the technical domain's AUC was significantly higher according to the exploratory classification analysis. These results show that different academic domains exhibit different predicting behaviour of LMS engagement metrics. Therefore, rather than using a standard method for engagement indicators, LMS-based predictive models used in adaptive OBE systems should be calibrated based on the characteristics of individual domains.