Table 5 DMEPOS mean and standard deviation of AUC with varying levels of RUS (10 iterations of fivefold cross-validation)
From: Evaluating classifier performance with highly imbalanced Big Data
Class ratio | LGB | CB-16 | ET | XGB-24 | RF-GPU-32 |
|---|---|---|---|---|---|
1:1 | 0.87070 | 0.91952 | 0.98862 | 0.97944 | 0.96163 |
(0.00145) | (0.00134) | (0.00059) | (0.00077) | (0.00082) | |
1:3 | 0.87610 | 0.92829 | 0.98874 | 0.98732 | 0.97472 |
(0.00161) | (0.00108) | (0.00072) | (0.00067) | (0.00082) | |
1:9 | 0.87922 | 0.92575 | 0.98774 | 0.98933 | 0.98175 |
(0.00169) | (0.00152) | (0.00095) | (0.00069) | (0.00060) | |
1:27 | 0.88011 | 0.91607 | 0.98536 | 0.98992 | 0.98511 |
(0.00165) | (0.00174) | (0.00111) | (0.00069) | (0.00066) | |
1:81 | 0.87140 | 0.90471 | 0.98207 | 0.98977 | 0.98549 |
(0.00241) | (0.00192) | (0.00124) | (0.00063) | (0.00074) | |
Unchanged | 0.85223 | 0.89143 | 0.97917 | 0.98959 | 0.98453 |
(0.00492) | (0.00207) | (0.00113) | (0.00069) | (0.00104) |