Table 2 Diagnostic performance of different machine learning models
From: Risk prediction and effect evaluation of complicated appendicitis based on XGBoost modeling
Outcome | Dataset | Xgboost | Random forest | CART | Support vector machine |
|---|---|---|---|---|---|
AUC | Training | 0.996(0.991,1.000) | 1.000(1.000–1.000) | 0.820(0.786–0.855) | 0.951(0.932–0.970) |
|  | Test | 0.914(0.874,0.955) | 0.870(0.819–0.921) | 0.742(0.671–0.813) | 0.882(0.833–0.931) |
Sensitivity(%) | Training | 95.9(94.3,97.5) | 100.0(100.0,100.0) | 72.8(69.2,76.4) | 94.4(92.5,96.3) |
| Â | Test | 86.5(81.7,91.3) | 88.8(84.4,93.2) | 71.9(65.6,78.2) | 86.5(81.7,91.3) |
Specificity(%) | Training | 97.4(96.1,98.7) | 99.4(98.8,100.0) | 86.9(84.2,89.6) | 88.5(85.9,91.1) |
| Â | Test | 84.6(79.5,89.7) | 69.2(62.7,75.7) | 75.0(68.9,81.1) | 77.9(72.0,83.8) |
Precision(%) | Training | 97.0(95.6,98.4) | 99.3(98.6,100.0) | 82.6(79.5,85.7) | 87.5(84.8,90.2) |
| Â | Test | 82.8(77.5,88.1) | 71.2(64.8,77.6) | 71.1(64.7,77.5) | 77.0(71.1,82.9) |
Accuracy(%) | Training | 96.7(95.3,98.2) | 99.7(98.8,100.0) | 80.3(76.9, 83.5) | 91.2(88.6, 93.4) |
| Â | Test | 85.5(80.5,90.5) | 78.2(71.7,83.8) | 73.6(66.8, 79.7) | 81.9(75.7, 87.0) |
PPV(%) | Training | 97.1(95.7,98.5) | 99.3(98.6,100.0) | 82.6(79.5,85.7) | 87.5(84.8,90.2) |
| Â | Test | 84.8(79.7,89.9) | 71.2(64.8,77.6) | 71.1(64.7,77.5) | 77.0(71.1,82.9) |
NPV(%) | Training | 96.6(95.1,98.1) | 1.00(100.0,100.0) | 78.8(75.5,82.1) | 94.9(93.1,96.7) |
| Â | Test | 89.7(85.4,94.0) | 87.8(83.2,92.4) | 75.7(69.6,81.8) | 87.1(82.4,91.8) |