Performances of the decision tree models for men (Tehran Lipid and Glucose Study 1999–2012)
| Types of decision tree models | Performance measures | C5.0 | QUEST | CART |
|---|---|---|---|---|
| Models (1) | Sensitivity | 72% | 78% | 74% |
| Specificity | 70% | 72% | 72% | |
| PPV | 23% | 25% | 24% | |
| NPV | 95% | 96% | 96% | |
| Accuracy | 71% | 73% | 72% | |
| F-Measure | 0.35 | 0.38 | 0.36 | |
| G-Mean | 0.71 | 0.75 | 0.73 | |
| AUC | 0.80 | 0.78 | 0.79 | |
| Models (2) | Sensitivity | 67% | 68% | 64% |
| Specificity | 73% | 78% | 74% | |
| PPV | 23% | 27% | 23% | |
| NPV | 95% | 95% | 94% | |
| Accuracy | 73% | 77% | 73% | |
| F-Measure | 0.34 | 0.39 | 0.34 | |
| G-Mean | 0.70 | 0.73 | 0.69 | |
| AUC | 0.74 | 0.77 | 0.78 |
Model (1) was developed based on 15 variables which included 2h-PCPG.
Model (2) was developed based on 14 variables (2h-PCPG was excluded).
F-Measure: Harmonic mean between PPV and sensitivity, F-Measure=2 (sensitivity×PPV)/(sensitivity+PPV).
G-Mean of sensitivity and specificity, G=√sensitivity×specificity.
2h-PCPG, 2-hour postchallenge plasma glucose; AUC, area under the curve; CART, Classification and Regression Tree; G-Mean, geometric mean; NPV, negative predictive value; PPV, positive predictive value; QUEST, Quick Unbiased Efficient Statistical Tree.