Skip to main content

Table 1 | Comparison of the performance of multiple prediction models

From: Clinical applications of machine learning in the survival prediction and classification of sepsis: coagulation and heparin usage matter

Methods

 

Accuracy

Precision

Recall

F1

AUC

Random Forest

 Training

0.851

1.000

0.238

0.384

0.619

 Test

0.808

0.909

0.068

0.127

0.533

Logistic Regression

 Training

0.825

0.629

0.256

0.364

0.610

 Test

0.808

0.567

0.260

0.357

0.605

Lasso Regression

 Training

0.825

0.762

0.148

0.248

0.568

 Test

0.813

0.710

0.151

0.249

0.567

Radial SVM [40]

 Training

0.515

0.970

0.491

0.652

0.701

 Test

0.337

0.896

0.204

0.333

0.586

 Val

0.806

0.849

0.920

0.883

0.642

Gradient boosting [40]

 Training

0.851

0.934

0.899

0.916

0.690

 test

0.718

0.822

0.816

0.819

0.574

 Val

0.828

0.885

0.905

0.895

0.682

Bayes [40]

 Training

0.567

0.965

0.553

0.703

0.649

 Test

0.465

0.861

0.405

0.551

0.562

 Val

0.828

0.891

0.895

0.893

0.713

Linear regression [40]

 Training

0.801

0.943

0.835

0.886

0.599

 Test

0.679

0.828

0.763

0.794

0.541

 Val

0.788

0.885

0.842

0.863

0.689

Linear SVM [40]

 Training

0.337

0.896

0.205

0.333

0.586

 Test

0.467

0.861

0.407

0.553

0.586

 Val

0.818

0.873

0.906

0.889

0.676

Sofa Score [13] DCQMFF (proposed)

 All data

0.752

0.371

0.327

0.348

0.807

 Training

0.822

0.822

0.821

0.822

0.896

 Test

0.821

0.812

0.812

0.812

0.885

 Val

0.775

0.764

0.754

0.759

0.849

CNN (Proposed)

 Training

0.928

0.924

0.856

0.888

0.953

 Test

0.924

0.887

0.845

0.865

0.947

 Val

0.834

0.825

0.818

0.821

0.909