Table 3 Training set performance of machine learning classifiers for prediction of wound outcome using all clinical and microbial metagenomic variables.
From: Metagenomic features of bioburden serve as outcome indicators in combat extremity wounds
Model | Probability threshold (%) | Median (%) | Mean (%) | q1 (%) | q3 (%) |
|---|---|---|---|---|---|
Precision | |||||
rf | 35.00 | 66.67 | 65.27 | 55.60 | 75.00 |
nnet | 40.00 | 57.14 | 57.84 | 50.00 | 66.70 |
glmnet | 45.00 | 60.00 | 59.85 | 50.00 | 66.70 |
svmRadial | 20.00 | 66.67 | 64.45 | 50.00 | 75.00 |
Sensitivity | |||||
rf | 35.00 | 83.33 | 80.15 | 71.40 | 90.00 |
nnet | 40.00 | 75.00 | 73.82 | 62.50 | 85.70 |
glmnet | 45.00 | 71.43 | 71.94 | 62.50 | 83.30 |
svmRadial | 20.00 | 66.67 | 66.76 | 50.00 | 83.30 |
Specificity | |||||
rf | 35.00 | 85.71 | 85.42 | 80.00 | 91.30 |
nnet | 40.00 | 82.97 | 82.16 | 76.40 | 87.50 |
glmnet | 45.00 | 84.62 | 83.93 | 78.90 | 90.00 |
svmRadial | 20.00 | 88.24 | 86.58 | 81.20 | 95.00 |
