Machine Learning Applications for Learning Early Warning System in Taiwan

機構典藏 > College of Business and Management > Graduate Institute & Department of Banking and Finance > Journal Article > Item 987654321/122916

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/122916

Title:	Machine Learning Applications for Learning Early Warning System in Taiwan
Authors:	Lin, Jyh-Jiuan;Tsai, Gwei-Hung;Chang, Ching-Hui
Keywords:	e-learning;e-portfolio;recommender system;learning management systems
Date:	2020-11-19
Issue Date:	2023-04-28 16:26:22 (UTC+8)
Abstract:	This research proposes an alternative approach reference for a learning early warning system implementation. Digital e-portfolio data of 6 semesters are used respectively to build 4 commonly used supervised machine learning (ML) classifiers including random forests (RF), support vector machine (SVM), extreme gradient boosting (XGBoost) and artificial neural networks (ANN). The empirical results from year 2013 to 2019 semesters, excluding 2018 due to sabbatical leave of the instructor, show that the top 2 classifiers are XGBoost and RF in terms of the following aggregated criteria consideration: 1. Accuracy, 2. Recall, 3. Precision, 4. F1-score, 5. AUC, 6. Cross-validation mean accuracy, 7. Crossvalidation accuracy standard deviation (StDev), and 8. Computation time. Since XGBoost has outperformed the rest classifiers, it is recommended to be deployed by the early warning system implementation. The evidence of the model robustness supports the approach of the learning early warning systems implementation incorporating ML methods. Besides, midterm score reaches a consensus for XGBoost and RF to be selected as the most significant features to identify at-risk students. Interestingly, the second most significant feature selected by RF is the “mock exam score”. It fits the purpose of mock exam which is designed to help students foresee the midterm test format. On the other hand, the second most significant feature selected by XGBoost is the “forum post counts” which implies that the higher the participation is, the better the academic performance gets empirically.
Relation:	Asian Journal of Information and Communications 12(1), p.77-89
Appears in Collections:	[Graduate Institute & Department of Banking and Finance] Journal Article

Files in This Item:

File	Description	Size	Format
Machine Learning Applications for Learning Early Warning System in Taiwan.pdf		502Kb	Adobe PDF	96	View/Open

Loading...