Table 2 Performance evaluation in binary stress detection.

Model	Precision (@Recall\(\approx\)0.9)	PR AUC	ROC AUC
Location MLP¹¹ GBDT⁴³ LSTM¹⁸ Clustered CALM-Net CALM-Net Branched CALM-Net	0.588 ± 0.002 0.849 ± 0.026 0.780 ± 0.004 0.817 ± 0.003 0.843 ± 0.003 0.845 ± 0.005	0.674 ± 0.012 0.877 ± 0.031 0.805 ± 0.010 0.877 ± 0.002 0.933 ± 0.001 0.931 ± 0.001	0.580 ± 0.011 0.582 ± 0.064 0.530 ± 0.020 0.684 ± 0.003 0.807 ± 0.002 0.805 ± 0.004
Transformer (Trans)^46,50 CATrans-Net Branched CATrans-Net	0.780 ± 0.004 0.851 ± 0.006 0.851 ± 0.004	0.786 ± 0.008 0.932 ± 0.002 0.933 ± 0.001	0.501 ± 0.015 0.805 ± 0.003 0.805 ± 0.002

The labels of median stress and very stressed are combined. We report the precision of different models for a recall of 0.9 and the Area Under the Curve (AUC) for the precision-recall curve and for the receiver operating characteristic (ROC). The PR and ROC curves are shown in Fig. 3.
[bold] values indicate top performers in each category.

Quick links

Search