Early Prediction of Sepsis Based on Machine Learning Algorithm

Table 8

Prediction performance of different features on the test set.

Variables	AUROC	Accuracy	Precision	Recall
Raw variables	0.9710	0.919	0.92	0.60
Raw variables + measurement intensity features	0.9756	0.927	0.93	0.64
Raw variables + window features	0.9767	0.928	0.93	0.64
Raw variables + medical indicator features	0.9718	0.919	0.92	0.60
Feature generation method features	0.9789	0.931	0.93	0.65