Table 2 Overall Mortality Prediction Performance
From: Improving medical machine learning models with generative balancing for equity and excellence
| Â | Downstream Method F1 Score | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Training Data | eICU | MIMIC | ||||||||||
| Â | Avg | KNN | LR | NN | RFC | XGB | Avg | KNN | LR | NN | RFC | XGB |
Original Data | 0.376 ± 0.02 | 0.180 ± 0.02 | 0.453 ± 0.03 | 0.486 ± 0.03 | 0.272 ± 0.02 | 0.490 ± 0.02 | 0.296 ± 0.00 | 0.112 ± 0.00 | 0.389 ± 0.00 | 0.428 ± 0.01 | 0.150 ± 0.00 | 0.400 ± 0.00 |
Upsampling | 0.381 ± 0.01 | 0.249 ± 0.02 | 0.485 ± 0.01 | 0.427 ± 0.02 | 0.240 ± 0.02 | 0.504 ± 0.01 | 0.130 ± 0.00 | 0.113 ± 0.01 | 0.170 ± 0.01 | 0.202 ± 0.01 | 0.011 ± 0.00 | 0.155 ± 0.00 |
Downsampling | 0.442 ± 0.01 | 0.273 ± 0.02 | 0.481 ± 0.01 | 0.481 ± 0.01 | 0.487 ± 0.01 | 0.486 ± 0.01 | 0.109 ± 0.00 | 0.053 ± 0.00 | 0.164 ± 0.00 | 0.150 ± 0.01 | 0.016 ± 0.00 | 0.161 ± 0.00 |
SMOTE | 0.427 ± 0.01 | 0.289 ± 0.02 | 0.461 ± 0.01 | 0.446 ± 0.01 | 0.440 ± 0.02 | 0.500 ± 0.01 | 0.252 ± 0.00 | 0.181 ± 0.01 | 0.353 ± 0.00 | 0.210 ± 0.01 | 0.124 ± 0.00 | 0.393 ± 0.00 |
MCRAGE | 0.376 ± 0.02 | 0.127 ± 0.01 | 0.454 ± 0.03 | 0.513 ± 0.02 | 0.295 ± 0.02 | 0.491 ± 0.03 | 0.298 ± 0.00 | 0.102 ± 0.00 | 0.389 ± 0.00 | 0.425 ± 0.00 | 0.172 ± 0.01 | 0.399 ± 0.00 |
SMA | 0.280 ± 0.01 | 0.111 ± 0.01 | 0.379 ± 0.01 | 0.451 ± 0.02 | 0.091 ± 0.02 | 0.369 ± 0.02 | – | – | – | – | – | – |
FairPlay | 0.457 ± 0.01 | 0.292 ± 0.02 | 0.498 ± 0.01 | 0.494 ± 0.01 | 0.487 ± 0.01 | 0.513 ± 0.01 | 0.329 ± 0.00 | 0.128 ± 0.00 | 0.401 ± 0.01 | 0.442 ± 0.01 | 0.266 ± 0.00 | 0.407 ± 0.01 |
