Table 4. Predictive performance of the six machine learning models in the validation set for post-RFCA recurrence.
| Model | AUC (95% CI) | Accuracy | F1 score | Sensitivity | PPV | NPV |
|---|---|---|---|---|---|---|
| Random forest | 0.781 (0.729–0.835) | 0.792 | 0.501 | 0.427 | 0.647 | 0.833 |
| SVM | 0.793 (0.782–0.805) | 0.804 | 0.489 | 0.381 | 0.710 | 0.825 |
| KNN | 0.702 (0.649–0.747) | 0.750 | 0.379 | 0.329 | 0.472 | 0.806 |
| Gradient boosting | 0.759 (0.661–0.837) | 0.762 | 0.425 | 0.363 | 0.545 | 0.814 |
| Decision tree | 0.616 (0.549–0.695) | 0.704 | 0.422 | 0.444 | 0.413 | 0.816 |
| XGBoost | 0.727 (0.655–0.800) | 0.754 | 0.448 | 0.410 | 0.502 | 0.821 |
RFCA, radiofrequency catheter ablation; AUC, area under the curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; SVM, support vector machine; KNN, k-nearest neighbor; XGB, extreme gradient boosting.