Table 4 Median internal AUROC differences (with interquartile range) across all prediction problems for each database and classifier when choosing the sampling strategy with the highest AUROC during CV
Database | Number of prediction tasks | Lasso | Random forest | XGBoost | All classifiers |
|---|---|---|---|---|---|
CCAE | 14 | − 0.0025 (0.0073) | 0.0001 (0.0106) | 0 (0.0067) | − 0.0004 (0.0076) |
MDCD | 17 | − 0.0004 (0.0044) | 0 (0.0062) | 0 (0.0071) | 0 (0.0068) |
MDCR | 19 | 0.0000 (0.0052) | 0.0037 (0.0143) | 0 (0.0057) | 0 (0.0075) |
IQVIA Germany | 8 | − 0.0011 (0.0048) | 0.0012 (0.0095) | − 0.0045 (0.0204) | − 0.0010 (0.0098) |
All databases | 58 | − 0.0004 (0.0053) | 0.0008 (0.0099) | 0 (0.0074) | 0 (0.0081) |