Predicting student success with self-regulated behaviors: A seven-year data analytics study on a Hong Kong University English Course

This study investigates how the online behaviors of students can be associated with student outcomes via a learning analytics approach. The objective of this study is to examine whether online behaviors of students can predict student success in an English course. Also, this study aims to examine whether some machine learning techniques performed better than others. Learners in this study were students in an English for Academic Purposes course at a university in Hong Kong. Other than face-to-face teaching, this course included a mandatory multimodal language learning package first introduced in 2012. The research team retrieved a mega-dataset from all learners who took the course from 2012 to 2019 (n = 17,968). Using the Classification Tree technique, Logistical Regression, and Artificial Neural Networks, the research team derived three predictive algorithms to associate various online engagement variables (e.g., start day, first attempt scores, and number of attempts) with course success. Results suggest that Classification Tree performed better in predictive accuracy and F1 ratio. The chapter provides useful insights for adaptive design. The study also discusses several useful suggestions on how practitioners can adopt these algorithms for better adaptive design. Copyright © 2024 selection and editorial matter, Patsy D. Moskal, Charles D. Dziuban, and Anthony G. Picciano; individual chapters, the contributors.