Using Ensemble Machine Learning and Feature Engineering to Increase the Accuracy of Predicting Learners' Performance in an Online Educational Environment
Background: Online training has gained popularity as an effective teaching method, necessitating diligent monitoring of learner progress and engagement. The challenge of predicting academic performance in online courses is crucial for supporting learners at risk of academic loss. This study aimed to...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Shiraz University of Medical Sciences
2024-12-01
|
| Series: | Interdisciplinary Journal of Virtual Learning in Medical Sciences |
| Subjects: | |
| Online Access: | https://ijvlms.sums.ac.ir/article_50587_e403e00e208e76873b73eab4816479b6.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Background: Online training has gained popularity as an effective teaching method, necessitating diligent monitoring of learner progress and engagement. The challenge of predicting academic performance in online courses is crucial for supporting learners at risk of academic loss. This study aimed to develop a robust model for predicting learners' performance using ensemble machine learning and feature engineering techniques.Methods: This research employed a classification approach based on the Digital Electronic Education and Design Suite (DEEDS) dataset, which records real-time interactions of learners within an online educational environment. The dataset analyzed in this research included activity logs from 115 undergraduate students majoring in computer engineering who participated in a digital electronics course at the University of Genoa, Italy, between September and December 2015. Various machine learning algorithms, including Random Forest (RF), Adaptive Boosting (AdaBoost), Gradient Boosting (GB), Light Gradient-Boosting Machine (LightGBM), and eXtreme Gradient Boosting (XGBoost), were applied. The study also utilized ensemble learning methods such as Boosting and Stacking to enhance prediction accuracy. Feature engineering techniques were implemented to extract and select relevant features from the dataset, leading to the development of a predictive model.Results: The proposed model achieved an accuracy of 97.43%, a precision of 96.20%, and an F1-score of 98.06%, indicating an acceptable predictive capability. Notably, the findings revealed that feature selection significantly enhanced performance; in the absence of feature selection, the accuracy dropped to 92.15%. Additionally, ensemble methods like Boosting and Stacking provided a 15% enhancement in prediction accuracy compared to traditional approaches. Overall, the integration of feature engineering and ensemble techniques acceptably optimized the model's ability to predict learners’ academic performance in online educational settings. Conclusion: This research validates the effectiveness of employing ensemble machine learning techniques and feature engineering in predicting learners’ academic performance in online education. Future studies should explore additional ensemble methods and incorporate diverse feature types to enhance prediction accuracy. |
|---|---|
| ISSN: | 2476-7263 2476-7271 |