TABLE 3.

Testing results of the predictive models: Stage 1 try-weaning model of the 60th HR and Stage 2 MV-weaning model of the 120th HR.

Algorithm	Accuracy	Sensitivity	Specificity	AUC
Stage 1
Logistic regression	0.710	0.710	0.710	0.776
Random forest	0.760	0.760	0.760	0.847
SVM	0.716	0.778	0.609	0.759
KNN	0.686	0.749	0.578	0.730
LightGBM	0.768	0.788	0.733	0.860
MLP	0.732	0.746	0.709	0.815
XGBoost	0.774	0.806	0.718	0.853
Stage 2
Logistic regression	0.827	0.827	0.826	0.913
Random forest	0.824	0.822	0.829	0.918
SVM	0.713	0.714	0.712	0.797
KNN	0.649	0.679	0.580	0.683
lightGBM	0.842	0.842	0.842	0.923
MLP	0.805	0.804	0.807	0.905
XGBoost	0.810	0.810	0.810	0.908