Acceso usuarios
Zhenfeng Jiang, Aidah Abdul Karim, Fariza Khalid
This study investigates student involvement in a blended learning course within Chinese higher education and assesses the potential of machine learning to facilitate ongoing engagement evaluation across time. Engagement is conceptualised within a dual-subject framework that examines the dynamic interplay between learners and instructors, and is implemented across behavioural, cognitive, and social-interaction dimensions. Data from 144 students over three semesters in an undergraduate course on Database Principles and Applications produced eleven indicators based on Learning Management Systems (LMS) activity logs and classroom behaviours. Predictive models were created utilising both batch learning and an incremental Random Forest, with “incremental” indicating a model that updates with fresh data without necessitating complete retraining. K-Means clustering was subsequently utilised to delineate unique engagement characteristics. The results indicate that the incremental model attained superior accuracy and demonstrated enhanced adaptability to newly arriving data. Clustering demonstrated diverse engagement patterns that underscore the importance of varied educational methodologies. Longitudinal observations demonstrated that modifications in instructional design positively influenced student engagement. This study enhances current LMS-based engagement research by integrating the dual-subject paradigm with progressively updated learning analytics. It also illustrates a pragmatic approach for educators to consistently assess engagement and implement timely, data-driven modifications in blended higher education settings
