Sign in Subscribe
Privacy

Engineering Researchers Develop Framework for Generating Synthetic Student Data for Learning Analytics

Strategic Insights

— 1 min read

Researchers at the University of Arkansas Little Rock have created a comprehensive framework for generating, evaluating, and utilizing synthetic student data for learning analytics, addressing the issue of sensitive student data and its associated privacy risks. By using two state-of-the-art synthesizers, the team was able to create synthetic datasets that mimic the statistical properties of real data without exposing personal information. The framework was tested on the Open University Learning Analytics Dataset (OULAD) and evaluated against real data in terms of machine learning utility, statistical quality, and privacy risk.

Key Takeaways:

  • The researchers employed two state-of-the-art synthesizers, the Gaussian Copula (GC) and Conditional Tabular GAN (CTGAN), to create synthetic versions of the Open University Learning Analytics Dataset (OULAD).
  • The synthetic data were rigorously evaluated against the original data along three critical axes: 1) machine learning utility for predicting student dropout and final grades, 2) statistical quality using the SDMetrics library, and 3) privacy risk via a membership inference attack.
  • CTGAN-generated data retained 93.1% of dropout-prediction AUC relative to real data while remaining robust to membership-inference attacks.
  • Gaussian Copula achieved higher statistical-fidelity scores but materially lower predictive utility, quantifying a practical utility-privacy trade-off for educational tabular synthesis.
  • The framework is designed to provide a promising solution by creating artificial datasets that mimic the statistical properties of real data without exposing personal information.
  • The researchers concluded that their framework has the potential to enhance educational outcomes by personalizing student support while addressing the issue of sensitive student data and its associated privacy risks.

Statistics:

  • 93.1% dropout-prediction AUC retained by CTGAN-generated data relative to real data.
  • 13(():177345-177351 - Volume and page numbers of Generative Private Synthetic Student Data for Learning Analytics: An Empirical Study publication.
  • 6287639 - IEEE Access publication number.
  • 10.1109/ACCESS.2025.3619091 - DOI for Generative Private Synthetic Student Data for Learning Analytics: An Empirical Study publication.

Sources:

  • Generative Private Synthetic Student Data for Learning Analytics: An Empirical Study. IEEE Access, 2025, 13():177345-177351.
  • https://doi-org.sdpl.idm.oclc.org/10.1109/ACCESS.2025.3619091
  • NewsRx LLC. University of Arkansas Little Rock Researchers Provide New Study Findings on Engineering (Generative Private Synthetic Student Data for Learning Analytics: An Empirical Study). Journal of Engineering. November 3, 2025; p 4804.

Read more