Table 2.
Comparison of top-4 machine learning models and logistic regression model in validation group.
| Models | Observed CI-AKI (n = 57) | Observed non-CI-AKI (n = 316) | Cut-off | ||
|---|---|---|---|---|---|
| True positive | False positive | False negative | True negative | ||
| RF10 | 42 | 15 | 91 | 225 | 0.50 |
| 73.7% | 26.3% | 28.8% | 71.2% | ||
| RF15 | 41 | 16 | 83 | 233 | 0.50 |
| 71.9% | 28.1% | 26.3% | 73.7% | ||
| RF20 | 37 | 20 | 84 | 232 | 0.50 |
| 64.9% | 35.1% | 26.6% | 73.4% | ||
| SVM15 | 41 | 16 | 106 | 210 | 0.50 |
| 71.9% | 28.1% | 33.5% | 66.5% | ||
| LR3 | 36 | 21 | 270 | 46 | 0.80 |
| 63.2% | 36.8% | 85.4% | 14.6% | ||
| Precision | Recall | F1-score | Specificity | Accuracy | |
| RF10 | 31.6% | 73.7% | 44.2% | 71.2% | 71.6% |
| RF15 | 33.1% | 71.9% | 45.3% | 73.7% | 73.5% |
| RF20 | 30.6% | 64.9% | 41.6% | 73.4% | 72.1% |
| SVM15 | 27.9% | 71.9% | 40.2% | 66.5% | 67.3% |
| LR3 | 11.8% | 63.2% | 19.8% | 14.6% | 22.0% |
RF10, random forest model with top 10 variables; RF15, random forest model with top 15 variables; RF20, random forest model with top 20 variables; SVM15, support vector machine model with top 15 variables; LR3, logistic regression model 3.