Cohort study demonstrates stable mid-semester classification of learner categories in university students, indicating strong potential for scalable personalized academic support.
Key Points
To develop and validate a transparent, scalable framework that uses real-time formative assessment data to categorize learners for targeted pedagogical recommendations.
Aggregated time-investment and performance metrics into two dimensions using separate principal component analyses.
Segmented component scores by medians to establish four learner categories, evaluating model stability through bootstrap analysis and cross-cohort transfer tests across three university cohorts.
Predictive accuracy steadily improved over the semester and stabilized following the third formative assessment around week 9, after which subsequent predictive gains diminished.
Component structures and prediction trajectories remained consistent across adjacent-year and two-year cohort transfers, with misclassifications predominantly clustered near median decision boundaries.