Table 2 Comparison of CV-AUC of standard random forest (RF) and tuned random forest (tRF). Screening is applied to both methods, but not class balancing or inverse sampling probability weighting
From: Improving random forest predictions in small datasets from two-phase sampling designs
RF | tRF | |
|---|---|---|
All markers | 0.824 | 0.807 |
T cell markers | 0.812 | 0.802 |
Antibody markers | 0.708 | 0.721 |
No markers | 0.442 | 0.455 |