The generalised plait ensemble outperforms single classifiers, majority-vote and stacking ensembles, substantially improving perceived task-difficulty classification (e.g., SVM plait error 14.54%, KNN plait 12.84%).¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2016 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? The best performing plait on average is the KNN plait, except for class hard where the SVM plait performed best; DCT plait showed weaker performance. → Ruth Janning 2016

Evidence¶

Ruth Janning 2016¶

Ruth Janning, Carlotta Schatten, Lars Schmidt-Thieme. (2016). Perceived Task-Difficulty Recognition from Log-file Information for the Use in Adaptive Intelligent Tutoring Systems. Int J Artif Intell Educ. https://doi.org/10.1007/s40593-016-0097-9

q2 · i? · design · r2

Per-student and per-class analysis of the same plait experiments (Fig. 2 and Fig. 3 plus Table 9). The figure-based analysis shows the plait "in each case the plait outperformed the other models, except for SVM and DCT applied to the data of student 7", and DCT plait "seems not always to work well".

"The best performing plait on average is the KNN plait, except for class hard where the SVM plait performed best. Overall, both SVM and KNN plait performed similarly good whereas DCT plait showed a weaker performance."

Discussion¶

Related Claims¶

• Log-file features are suitable for recognising perceived task-difficulty per task, with SVM the best single classifier (25.24% average test error). — 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
• Ensemble modeling of multiple LLM scorers yields performance gains in automated divergent thinking scoring — a broader claim this one bears on
• Rank-averaged ensembling stabilized predictions but did not outperform the single base model — related
• A Soft-Voting Ensemble achieves the best behavioral-dataset attrition prediction (CV F1 = 0.737, test recall = 0.769) — related
• Single log-file features per task show very small p-values but weak fit, with number of hints seen the strongest single predictor of perceived task-difficulty. — related