Machine Learning–Based Assessment of Student Engagement in Blended Learning: A Case Study in Chinese Higher Education

Evaluación del compromiso estudiantil en el aprendizaje combinado mediante aprendizaje automático: un estudio de caso en la educación superior china

Abstract

This study investigates student engagement in a blended learning environment within Chinese higher education, aiming to explore how machine learning techniques can support engagement assessment over time. The research focuses on an undergraduate course titled Database Principles and Applications, where engagement is examined through behavioral, cognitive, and social interaction dimensions. A total of 144 students participated across three semesters. Eleven indicators were developed based on students’ online activity logs and classroom behaviors. The study employed both batch and incremental learning methods to build predictive models. Specifically, the performance of an incremental Random Forest algorithm was compared with that of traditional batch approaches. In addition, K-Means clustering was used to identify distinct engagement profiles among students. The results showed that the incremental model offered higher accuracy and better adaptability to new data. Clustering analysis revealed diverse participation patterns, suggesting the need for differentiated instructional strategies. Longitudinal observations also indicated that changes in teaching design—such as task restructuring and more integrated learning activities—had a positive effect on student engagement. These findings suggest that data-driven methods may help instructors monitor engagement continuously and make timely instructional adjustments. Overall, the study demonstrates a practical approach to using machine learning for engagement tracking in blended learning and provides insights that may inform teaching strategies in similar higher education contexts.