Deep learning models (LSTM, GRU, transformer) achieve high and comparable accuracy (0.90–0.93) on physiological exam-outcome prediction¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r1 · 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? LSTM achieved a mean accuracy of 0.9299, GRU 0.9185, and the adapted transformer 0.9039, with the transformer performing comparably to the recurrent models. → Lala Yamazaki and Ramchandra Rimal 2026

Evidence¶

Lala Yamazaki and Ramchandra Rimal 2026¶

Lala Yamazaki and Ramchandra Rimal. (2026). Leveraging Physiological Signals to Predict Exam Outcomes with Machine Learning. https://arxiv.org/abs/2606.14960 doi:10.48550/arxiv.2606.14960

q2 · i? · associational · r1

Test-set evaluation over 30 replicates of tuned deep learning models; the article reports GRU and LSTM "mean performance metrics of 0.92 and 0.93" and Table 5 prints accuracies of 0.9185 (GRU) and 0.9299 (LSTM).

"The GRU and LSTM models yield reliable results, with mean performance metrics of 0.92 and 0.93, respectively, demonstrating excellent generalization to unseen data."

Lala Yamazaki and Ramchandra Rimal 2026 (2)¶

Lala Yamazaki and Ramchandra Rimal. (2026). Leveraging Physiological Signals to Predict Exam Outcomes with Machine Learning. https://arxiv.org/abs/2606.14960

q2 · i? · associational · r1

Same comparative test evaluation; the article reports the transformer's "slightly lower mean metric of approximately 0.90" (Table 5: accuracy 0.9039 ± 0.0081), describing it as versatile on tabular data though slightly below GRU and LSTM.

"The transformer model has a slightly lower mean metric of approximately 0.90, but it shows versatility by obtaining dependable performance in modeling the tabular data."

Discussion¶

Related Claims¶

• Random forest achieves the highest mean test accuracy (0.9590) of all models compared for predicting exam grade categories from physiological signals — reports the opposite
• Pairwise statistical tests confirm random forest's performance differences from other models are statistically significant (p < 0.05) — reports the opposite
• Logistic regression performs poorly (mean accuracy 0.4243) and SVM moderately (0.7374) on physiological exam-outcome prediction — related
• Using all collected physiological features improves exam-grade prediction accuracy to 96% over the 80% achieved by prior EDA-only analyses — related
• Random forest feature importance identifies EDA, ACC, and TEMP as the top three physiological predictors of exam outcomes — related
• The DMO-GRU model achieves the best classification results among the compared models, with 91.2% accuracy, 90.8% F1-score, 90.1% precision and 92.5% recall — related
• GRU is the lowest-performing model across both single-task and multi-task competency modeling, likely due to its two gating units versus LSTMs' three — reports the opposite