k-nearest neighbours achieves the highest recall of the tested models, making it better suited when identifying which students know a topic matters¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? KNN scores highest for recall, capturing more students who get a question correct, and consequently has the highest F1 despite a higher log loss than BMM and multi-question logistic regression. → Benjamin Mawdsley 2026

Evidence¶

Benjamin Mawdsley 2026¶

Benjamin Mawdsley, Tom Quilter, Richard Turner, Sarah Jackson, and Paul Edwards. (2026). Archetypes or ability? Clustering for modelling student mathematical competence. https://arxiv.org/abs/2607.26063 doi:10.48550/arxiv.2607.26063

q2 · i? · associational · r2

Model comparison in the predictive evaluation (Table 1: KNN recall 0.769±0.014 versus LogRegm 0.759±0.014 and BMM 0.754±0.015); the authors suggest KNN "may be the better suited model" when prioritising topics to revise.

"Interestingly, the KNN approach scores highest for recall, meaning that it captures more of the students who get a question correct. In cases where capturing which students know a topic is important, such as when prioritising which topics to revise, it may be the better suited model."

Discussion¶

Related Claims¶

• 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
• CatBoost achieved the best classification performance among tested algorithms for predicting course completion risk (F-measure .77, accuracy 78%, AUC .87) — related
• A stacking ensemble of LG, SVM, NB, KNN and XGB achieved the best fine-tuning F1-score (0.88), outperforming individual models such as SVM (0.86) — related