Table 3 Diagnostic performance of four machine learning algorithms
Machine learning algorithms | Models | Sensitivity | Specificity | Non-error Rate | AUC (95% CI) |
|---|---|---|---|---|---|
Support vector machine | Metabolites | 0.878 | 0.837 | 0.846 | 0.887 (0.857–0.915) |
Risk factors | 0.956 | 0.909 | 0.919 | 0.979 (0.971–0.986) | |
Random forest | Metabolites | 0.956 | 0.971 | 0.968 | 0.993 (0.988–0.997) |
Risk factors | 0.995 | 0.996 | 0.996 | 1.000 (0.999–1.000) | |
K-nearest neighbor | Metabolites | 0.985 | 0.734 | 0.785 | 0.914 (0.895–0.930) |
Risk factors | 0.980 | 0.871 | 0.893 | 0.977 (0.970–0.984) | |
Logistic regression | Metabolites | 0.673 | 0.742 | 0.728 | 0.755 (0.715–0.794) |
Risk factors | 0.847 | 0.912 | 0.861 | 0.944 (0.929–0.959) |