Skip to main content

Table 4 Results from PRBF algorithm from experts D1-D4. A) Cross-validation model training results for PRBF algorithm for Population size = 4000, stretch = 25, learning rate = 0.1, and training iterations = 100,000, B) True labels and predicted uncertain labels for the tested SFRP sample of fictitious patient number 72

From: Toward a hemorrhagic trauma severity score: fusing five physiological biomarkers

A
  Training accuracy Test accuracy
Fold-1 0.95 0.90
Fold-2 0.96 0.90
Fold-3 0.98 0.95
Fold-4 0.95 0.95
Fold-5 0.94 0.90
Mean accuracy 0.96 0.92
Standard deviation ± 0.01 ± 0.03
B
Fictitious patient Majority vote Uncertain label (\({\text{u}}\)) PRBF prediction (\(\pi\))
72 0 [1,0.5,0,0,0] [0.979,0.321,0,0,0]