Table 4 Test-set performance of DNN models using different strategies to address class imbalance.
From: Transfer learning prediction of type 2 diabetes with unpaired clinical and genetic data
Model | Sensitivity | Specificity | Balanced acc. | AUC |
|---|---|---|---|---|
SMOTE-NC | 0.2105 | 0.9705 | 0.5905 | 0.8263 |
Standard BCE | 0.0727 | 0.9979 | 0.5353 | 0.8178 |
pw_BCE | 0.8170 | 0.6941 | 0.7556 | 0.8275 |
WRS | 0.6541 | 0.8587 | 0.7564 | 0.8458 |
pw_BCE + WRS | 0.7419 | 0.7835 | 0.7627 | 0.8477 |
