TABLE II.
Validation AUC of the tumor cohort across different classifiers
| Classifier | Augmented variables | Traditional variables | ||
|---|---|---|---|---|
| AUC | SD | AUC | SD | |
| Random Forest | 0.826 | 0.010 | 0.774 | 0.012 |
| Lasso | 0.810 | 0.011 | 0.780 | 0.012 |
| Naïve bayes | 0.794 | 0.011 | 0.757 | 0.012 |
| Neural nets | 0.772 | 0.012 | 0.481 | 0.014 |
| kNN | 0.767 | 0.012 | 0.712 | 0.014 |
| SVM | 0.761 | 0.011 | 0.667 | 0.013 |
| Logistic regression | 0.563 | 0.053 | 0.727 | 0.014 |
Table II shows AUC and bootstrapped standard deviation of the tumor validation dataset for the best performing combination of tuning parameters and sampling techniques for each family of classifiers