Comparison between machine learning methods for mortality prediction for sepsis patients with different social determinants

Table 2 Detailed performances on the entire testing set

	Accuracy	AUC	Precision	Recall	F1_binary	F1_macro	Specificity
Ridge classifier	0.6790	0.7774	0.2682	0.7052	0.3886	0.5855	0.6745
Perceptron	0.6720	0.7786	0.2634	0.7052	0.3835	0.5801	0.6664
Passive-aggressive	0.6841	0.7582	0.2733	0.7131	0.3951	0.5907	0.6792
kNN	0.7135	0.7299	0.2780	0.6135	0.3826	0.5981	0.7305
Random forest	0.7516	0.6459	0.2826	0.4661	0.3519	0.5991	0.7999
LinearSVC_L1	0.6749	0.7781	0.2654	0.7052	0.3856	0.5823	0.6698
LinearSVC_L2	0.6784	0.7777	0.2678	0.7052	0.3882	0.5850	0.6739
SGDClassifier_L1	0.6790	0.7759	0.2682	0.7052	0.3886	0.5855	0.6745
SGDClassifier_L2	0.6790	0.7749	0.2668	0.6972	0.3859	0.5843	0.6759
SGDClassifier_EN	0.6801	0.7753	0.2683	0.7012	0.3881	0.5858	0.6765
MultinomialNB	0.6392	0.7040	0.2348	0.6614	0.3466	0.5487	0.6354
BernoulliNB	0.3107	0.5724	0.1665	0.9402	0.2830	0.3096	0.2042
Logistic regression	0.6824	0.7761	0.2720	0.7131	0.3938	0.5893	0.6772
SVC_rbf	0.6847	0.7744	0.2702	0.6932	0.3888	0.5882	0.6833
SVC_poly	0.6749	0.7751	0.2654	0.7052	0.3856	0.5823	0.6698
SVC_sigmoid	0.6277	0.6873	0.2349	0.6972	0.3514	0.5451	0.6159

F1 binary: F1 score for the positive class; F1_macro: macro-averaged F1 score; Passive-aggressive: passive-aggressive classifier; kNN: k-Nearest Neighbors; LinearSVC_L1 or _L2: support vector machine with linear kernel coupled with L1 or L2 regularization; SGDClassifier_L1 or _L2 or _EN: stochastic gradient descent with L1 or L2 or Elastic Net regularization; MultinomialNB: Multinomial naïve Bayes; BernoulliNB: Bernoulli naïve Bayes; SVC_rbf or _poly or _sigmoid: support vector machine with rbf kernel or polynomial kernel or sigmoid kernel

ISSN: 1472-6947