Table 3 Comparison of model performance in the hold-out testing dataset.
From: Predicting radiocephalic arteriovenous fistula success with machine learning
Model | AUROC | AUPRC | Accuracy | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|---|---|
Elastic Net | 0.807 | 0.737 | 71.3% | 56.4% | 83.0% | 72.1% | 70.9% |
Lasso | 0.794 | 0.719 | 72.5% | 66.7% | 77.0% | 69.3% | 74.8% |
Random Forest | 0.791 | 0.699 | 69.1% | 66.7% | 71.0% | 64.2% | 73.2% |
Logistic Regression | 0.786 | 0.729 | 73.6% | 66.7% | 79.0% | 71.2% | 75.2% |
Boosted Trees | 0.779 | 0.697 | 70.2% | 67.9% | 72.0% | 65.4% | 74.2% |
Pruned Tree | 0.730 | 0.586 | 66.9% | 62.8% | 70.0% | 62.0% | 70.7% |
UAB | — | — | 65.2% | 78.2% | 55.0% | 57.5% | 76.4% |
KDOQI | — | — | 61.8% | 23.1% | 92.0% | 69.2% | 60.5% |
