Variable selection methods for developing a biomarker panel for prediction of dengue hemorrhagic fever

BMC Research Notes

Table 1 Model assessment of the predictive power

Predictive	Accuracy	Accuracy	Accuracy	AUC	AUC	AUC
Models	Train	Test	∆	Train	Test	∆
Bagging Ensemble	0.804	0.745	0.059	0.994	0.704	0.29
GPS	0.882	0.804	0.078	0.976	0.921	0.055
MARS	0.902	0.726	0.176	0.955	0.789	0.166
TreeNet Gradient Boosting	0.804	0.745	0.059	0.994	0.704	0.29

Shown are the predictive powers for training and test samples based on a 10-fold cross validation procedure.
∆, difference between the training sample and the test sample; AUC, area under the curve.

ISSN: 1756-0500