Improving random forest predictions in small datasets from two-phase sampling designs

Table 4 Comparison of CV-AUCs of four stacking models and two random forest models

	CV-AUC
RF: T cell markers + GLM: antibody markers	0.838
RF: T cell markers + GLM: All markers	0.831
RF: T cell markers	0.819
RF: All markers + GLM: ntibody markers	0.821
RF: All markers + GLM: All markers	0.821
RF: All markers	0.824

Screening is applied for both GLM and RF, but class balancing is not done for RF. Inverse sampling probability weights are used in both GLM and RF training. Clinical covariates are included in the predictors of all RF and GLM models

ISSN: 1472-6947