Development of early prediction model of in-hospital cardiac arrest based on laboratory parameters

Table 3 Summary of model performance of five machine learning algorithms

Model	AUC	Accuracy	Specificity	Sensitivity	F1-score
Training set
Extra trees classifier	0.939	0.858	0.829	0.901	0.863
Logistic regression	0.896	0.806	0.840	0.753	0.794
Random forest	0.938	0.853	0.825	0.894	0.858
Support vector machine	0.829	0.511	0.504	0.987	0.667
Decision tree	0.871	0.825	0.783	0.896	0.836
Testing set
Extra trees classifier	0.920	0.834	0.818	0.869	0.843
Logistic regression	0.895	0.807	0.841	0.766	0.802
Random forest	0.877	0.784	0.831	0.725	0.774
Support vector machine	0.864	0.542	0.527	0.995	0.689
Decision tree	0.843	0.794	0.768	0.856	0.809
Fivefold cross-validation
Extra trees classifier	0.920	0.841	0.787	0.895	0.849
Logistic regression	0.894	0.808	0.816	0.798	0.805
Random forest	0.889	0.803	0.853	0.754	0.792
Support vector machine	0.867	0.570	0.156	0.985	0.697
Decision tree	0.848	0.806	0.753	0.858	0.816

ISSN: 1475-925X