Table 2.
Comparative accuracy of six machine learning algorithms in predicting SARS-CoV-2 asymptomatic status from immunological factors.
| Machine learning model | Pseudo-R2 (10% evaluation holdback sample 1) (%) | Pseudo-R2 (10% evaluation holdback sample 2) (%) | Average Pseudo-R2 (%) | |
| With SCGF-β a | ||||
|
|
Decision tree | 100.00 | 100.00 | 100.00 |
|
|
XGBoostb | 100.00 | 100.00 | 100.00 |
|
|
GLMc (logistic) | 100.00 | 98.89 | 99.45 |
|
|
Random forest | 99.46 | 94.83 | 97.15 |
|
|
SVMd | 78.81 | 96.99 | 87.90 |
| Without SCGF-β | ||||
|
|
Random forest | 97.68 | 91.91 | 94.80 |
|
|
GLM (logistic) | 100.00 | 85.96 | 92.98 |
|
|
SVM | 77.76 | 89.69 | 83.73 |
|
|
XGBoost | 99.42 | 54.27 | 76.85 |
|
|
Decision tree | 100.00 | 2.22 | 51.11 |
aSCGF-β: stem cell growth factor-beta.
bXGBoost: extreme gradient boosting.
cGLM: generalized linear model.
dSVM: support vector machine.