A calibrated logistic-regression model predicts next-assessment pass/fail strongly on OULAD at both decision horizons¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? The calibrated logistic-regression baseline achieves AUC 0.884 at d-14 and 0.889 at d-7 on held-out test presentations. → Ngoc Luyen Le 2026 q2 i? Brier scores of 0.128 (d-14) and 0.122 (d-7) indicate usable calibration for downstream recourse generation. → Ngoc Luyen Le 2026 (2)

Evidence¶

Ngoc Luyen Le 2026¶

Ngoc Luyen Le, Marie-Hélène Abel, Bertrand Laforge. (2026). From Student Risk Prediction to SC2R: Semantics-Constrained Counterfactual Recourse for Educational Decision Support. arXiv preprint. https://arxiv.org/abs/2608.17618 doi:10.48550/arxiv.2608.17618

q2 · i? · design · r2

Offline evaluation on OULAD with a frozen module-presentation-level split (15 train, 2 validation, 5 test presentations), reporting held-out test performance. AUC values are 0.884 (d-14) and 0.889 (d-7); the article notes that "the later snapshot benefits from more recent learner evidence".

"The calibrated logistic-regression baseline provides strong perfor- mance, with AUC values of0.884atd−14and0.889atd−7. F1 and average precision follow the same pattern, indicating that the later snapshot benefits from more recent learner evidence."

Ngoc Luyen Le 2026 (2)¶

Ngoc Luyen Le, Marie-Hélène Abel, Bertrand Laforge. (2026). From Student Risk Prediction to SC2R: Semantics-Constrained Counterfactual Recourse for Educational Decision Support. arXiv preprint. https://arxiv.org/abs/2608.17618

q2 · i? · design · r2

Same offline OULAD evaluation of the calibrated logistic-regression pipeline. The printed Brier scores are 0.128 and 0.122, which the article says "indicate usable calibration for downstream recourse generation"; no effect size is reported.

"F1 and average precision follow the same pattern, indicating that the later snapshot benefits from more recent learner evidence. The Brier scores are0.128and0.122, which indicate usable calibration for downstream recourse generation."

Discussion¶

Related Claims¶

• Integer-programming recourse generates compact, model-valid, SHACL-conformant intervention plans at scale on OULAD — related
• Supplementary nonlinear models yield only modest AUC gains over the calibrated logistic baseline — related
• Baseline LR and XGBoost models achieve comparable predictive performance on OULAD, with XGBoost slightly higher AUC-ROC — related
• Explainable models including logistic regression, kNN and a Bernoulli Mixture Model all reach around 78% accuracy in predicting question-level pass/fail, with well-calibrated probabilities — related
• Machine learning methods outperformed logistic regression in classification accuracy across reviewed studies — reports the opposite