Table 2. Classification Models and EML Diagnostic Performance.
| Enrollment | Classification model | % (SE) | Likelihood ratio | Accuracy | ||||
|---|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | PPV | NPV | Positive | Negative | |||
| Training set | Decision tree | 95.0 (3.4) | 97.5 (1.7) | 95.0 (3.4) | 0.97.5 (1.7) | 38.0 | 0.05 | 96.7 |
| Naive Bayes | 65.0 (7.5) | 96.3 (2.1) | 89.7 (5.7) | 84.6 (3.8) | 17.3 | 0.36 | 85.8 | |
| Random forest | 87.5 (5.2) | 100.0 (0.0) | 100.0 (0.0) | 94.1 (2.6) | ND | 0.13 | 95.8 | |
| k–Nearest neighbors | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | ND | 0.00 | 100.0 | |
| Artificial neural network | 92.5 (4.2) | 100.0 (0.0) | 100.0 (0.0) | 96.4 (2.0) | ND | 0.08 | 97.5 | |
| Linear discriminant analysis | 50.0 (7.9) | 100.0 (0.0) | 100.0 (0.0) | 80.0 (4.0) | ND | 0.50 | 83.3 | |
| Support vector machine | 55.0 (7.9) | 100.0 (0.0) | 100.0 (0.0) | 81.6 (3.9) | ND | 0.45 | 85.0 | |
| Linear regression | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | ND | 0.00 | 100.0 | |
| Deep learning | 97.5 (2.5) | 98.8 (1.2) | 97.5 (2.5) | 98.8 (1.2) | 78.0 | 0.03 | 98.3 | |
| Partial least squares–discriminant analysis | 92.5 (4.2) | 100.0 (0.0) | 100.0 (0.0) | 96.4 (2.0) | ND | 0.08 | 97.5 | |
| EML | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | 100.0 (0.0) | ND | 0.00 | 100.0 | |
| Test set | EML | 100.0 (0.0) | 99.9 (1.0) | 88.9 (7.4) | 100.0 (0.0) | 707.0 | 0.0 | 99.9 |
Abbreviations: EML, ensemble machine learning; ND, not determinable; NPV, negative predictive value; PPV, positive predictive value.