Table 2.
Predictive performance (AUC) of 8 models and using several variables alone.
| Model | AUC | No. of optimal dimensions | ||
|---|---|---|---|---|
| Mean | SD | 95% CI | ||
| AdaBoost | 0.873 | 0.048 | 0.779–0.968 | 7 |
| ANN | 0.868 | 0.049 | 0.772–0.964 | 7 |
| DT | 0.802 | 0.057 | 0.691–0.913 | 2 |
| GBDT | 0.899 | 0.044 | 0.813–0.985 | 11 |
| LR | 0.867 | 0.049 | 0.771–0.963 | 13 |
| MNB | 0.784 | 0.058 | 0.670–0.898 | 11 |
| RFC | 0.890 | 0.045 | 0.801–0.979 | 13 |
| XGBoost | 0.883 | 0.047 | 0.792–0.975 | 7 |
| Tumor size | 0.753 | 0.023 | 0.707–0.798 | 1 |
| SUVmax | 0.734 | 0.024 | 0.688–0.780 | 1 |
| CEA | 0.720 | 0.026 | 0.669–0.770 | 1 |
AUC, area under the receiver operating characteristic curve; AdaBoost, adaptive boosting; ANN, artificial neural network; DT, decision tree; GBDT, gradient boosting decision tree; LR, logistic regression; MNB, multinomial naïve Bayes; RFC, random forest classifier; XGBoost, extreme gradient boosting; SUVmax, maximal standardized uptake value; CEA, carcinoembryonic antigen; CA125, carbohydrate antigen 12-5; Cyfra211, cytokeratin 19-fragments; CA153, carbohydrate antigen 15-3.