Table 4 AUC comparison after hyperparameter optimisation* between NOTE and cWGAN with rSVD on the HE and GMSC datasets. Benchmarks12 for cWGAN oversampling without extraction (best AUC highlighted in bold).
From: NOTE: non-parametric oversampling technique for explainable credit scoring
HE | NOTE | rSVD | rSVD and NSA | Engelmann and Lessmann12 | GMSC | NOTE | rSVD | rSVD and NSA | Engelmann and Lessmann12 |
|---|---|---|---|---|---|---|---|---|---|
LR | 0.9595 | 0.8152 | 0.9257 | 0.7554 | LR | 0.9750 | 0.9562 | 0.9643 | 0.7449 |
DT | 0.9546 | 0.8682 | 0.9227 | 0.7935 | DT | 0.9648 | 0.9642 | 0.9641 | 0.6090 |
ET | 0.9913 | 0.9777 | 0.9804 | N/A | ET | 0.9831 | 0.9735 | 0.9776 | N/A |
RF | 0.9891 | 0.9685 | 0.9775 | 0.9761 | RF | 0.9837 | 0.9806 | 0.9816 | 0.8412 |
GB | 0.9891 | 0.9733 | 0.9803 | 0.9119 | GB | 0.9834 | 0.9822 | 0.9829 | 0.8304 |
