EUVIMEDEUVIMED entdecken
Crossref

Behavioural engagement as an early predictor of noncompletion in online learning: A machine learning approach

Dung Hai Dinh, Phuong Linh Do

Advances in Online Education: A Peer-Reviewed Journal · 01.06.2026 · 4 (4) · 290

DOI: 10.69554/mxdh4725 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Advances in Online Education: A Peer-Reviewed Journal, online erschienen am 01.06.2026. Verfasst von Dung Hai Dinh und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Even as online course enrolments rise, completion rates have remained persistently low. This study examines the early prediction of learner noncompletion or dropout in online education by converting behavioural engagement signals into actionable risk indicators. In response to persistent attrition and its strong association with learner motivation, the study implements and evaluates machine learning models (ML) that prioritise observable activity patterns rather than demographic characteristics. Using a public Massive Open Online Course (MOOC) dataset comprising 416,920 learners from MITx and HarvardX, the data were cleaned, standardised, and reduced through correlation-based feature selection. Four behavioural indicators, namely active days, interaction frequency, content exploration, and chapters engaged, were retained, while outcome-adjacent variables were excluded to prevent information leakage. Five models (Logistic Regression, Decision Tree, Random Forest, Naïve Bayes, and Support Vector Machine [SVM]) were trained on a balanced dataset and evaluated using accuracy, precision, recall, F1-score, and Receiver Operating Characteristic – Area Under Curve (ROC-AUC). Logistic Regression achieved the highest overall accuracy (0.9797) and ROC-AUC (0.9898), while Random Forest demonstrated the most balanced performance (precision = 0.7106, recall = 0.7220, F1 = 0.7163). Naïve Bayes and SVM produced very high recall (approximately 0.96) but lower precision (around 0.42), indicating their suitability for broad early warning screening where minimising false negatives is critical. The results highlight behavioural engagement as the primary driver of learner persistence and suggest that theoretically grounded, interpretable models can support real-time early-warning dashboards for targeted intervention strategies. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.

Zitieren

APA

Dinh, D. H., & Do, P. L. (2026). Behavioural engagement as an early predictor of noncompletion in online learning: A machine learning approach. Advances in Online Education: A Peer-Reviewed Journal, 4(4), 290. https://doi.org/10.69554/mxdh4725

Vancouver

Dinh DH, Do PL. Behavioural engagement as an early predictor of noncompletion in online learning: A machine learning approach. Advances in Online Education: A Peer-Reviewed Journal. 2026;4(4):290. doi:10.69554/mxdh4725

BibTeX
@article{Dinh2026,
  author = {Dinh, Dung Hai and Do, Phuong Linh},
  title = {Behavioural engagement as an early predictor of noncompletion in online learning: A machine learning approach},
  journal = {Advances in Online Education: A Peer-Reviewed Journal},
  year = {2026},
  volume = {4},
  number = {4},
  pages = {290},
  doi = {10.69554/mxdh4725}
}
Kostenlos starten →