Digital Learning Engagement, Learner Diversity, and Academic Achievement in Secondary Education: A Quantitative Analysis of LMS Interaction Data.

Abstract

Digital learning contexts produce behavioural data that could inform more accurate understanding of students' learning and engagement, but often these behavioural data are studied in isolation with less consideration of learner diversity and context. With a cross sectional, explanatory-predictive design, this study examined the relationship among digital learning engagement, learner characteristics, and academic achievement in secondary education. A total of 281 middle and high school records from the xAPI-Edu-Data dataset were analysed with engagement defined as the number of times students raised their hand, visited a resource, viewed an announcement, and participated in discussion. The engagement index was standardised and tested for validity using non-parametric tests and ordinal logistic regression, and compared with multinomial logistic regression, random forest and gradient boosting models by repeated stratified five times cross validation. There were positive and consistent associations for digital engagement with the low, middle and high achievement groups, and a strong positive association with ordered achievement. The odds of higher achievement levels were about 6.7 times greater for a one-SD increase in engagement in the adjusted ordinal model. The best predictive model was the combination of random forest models, which had a macro area under the curve of 0.928 and an accuracy of about 79.5%. The findings show that combining LMS engagement and attendance, parent and learner-context factors offers a stronger foundation for achievement classification and data-informed student support than behavioural factors alone.