Global and adaptive split conformal prediction for capacity-constrained Top-K screening: a methodological evaluation using synthetic student performance data

A study in 'Frontiers in Education' introduces a methodological workflow designed to improve the utility of learning analytics models, particularly when review capacity is limited. It addresses the common issue of predictive scores lacking clear uncertainty quantification for prioritization. The proposed workflow integrates predictive benchmarking, adaptive split conformal calibration, and analytical routing, demonstrating that regularized linear regression yielded the most favorable outcomes when applied to synthetic student performance data.

Why it matters

This research provides a framework for enhancing the practical application of predictive analytics, particularly in resource-constrained environments where accurate prioritization is critical. By integrating uncertainty quantification, it enables more informed decision-making and resource allocation based on model predictions, improving the efficiency and effectiveness of screening and intervention processes.

Key insights

• Learning analytics models frequently generate predictive scores without explicit uncertainty measures to guide prioritization, especially under capacity constraints.
• A novel methodological workflow was developed, incorporating predictive benchmarking, global and normalized adaptive split conformal calibration, hypothetical capacity-constrained Top-K screening, threshold-relative analytical routing, and deterministic audit recording.
• The workflow was evaluated using a public synthetic student performance dataset comprising 6,607 records, 19 mixed-type predictors, and a numerical examination score outcome.
• Seven distinct regression models were compared across five repeated train-calibration-test partitions (60%/20%/20%).
• Regularized linear regression emerged as the most effective model, demonstrating superior performance within the evaluated methodology.

Source

Frontiers in Education — https://www.frontiersin.org/articles/10.3389/feduc.2026.1871887