Date
Publisher
arXiv
This study proposes and validates a Federated Learning (FL) framework to proactively identify at-risk students while preserving data privacy. Persistently high dropout rates in distance education remain a pressing institutional challenge. Using the large-scale OULAD dataset, we simulate a privacy-centric scenario where models are trained on early academic performance and digital engagement patterns. Our work investigates the practical trade-offs between model complexity (Logistic Regression vs. a Deep Neural Network) and the impact of local data balancing. The resulting federated model achieves strong predictive power (ROC AUC approximately 85%), demonstrating that FL is a practical and scalable solution for early-warning systems that inherently respects student data sovereignty.
Study at a glance
- Question: Can decentralized machine learning predict at-risk students as accurately as traditional centralized models while protecting student privacy?
- Setting and data: UK Open University, 22,437 distance education students across seven course modules, analyzing digital engagement and assessment data.
- Method: Machine learning model evaluation comparing centralized algorithms and federated learning simulations across course modules.
Why it matters
High dropout in distance education is an institutional challenge, but centralizing sensitive student data creates major privacy risks. Frameworks are needed that enable predictive insights without compromising data privacy and student data sovereignty.
Key findings
- Decentralized federated learning achieved predictive performance nearly equal to centralized models, reaching an area under the ROC curve of 84% to 85% compared to 86% centralized.
- Applying local data balancing before decentralized training improved student failure recall while achieving an 85% area under the ROC curve.
- Higher early assessment scores and greater virtual learning engagement were both associated with a lower risk of course failure.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who age?
Why use AI?
Study design
