Skip to main content

Table 6 Evaluation results of different processing methods under different scenarios of MAR (the ratio of missing proportion 2:1) mechanism

From: Application of machine learning missing data imputation techniques in clinical decision making: taking the discharge assessment of patients with spontaneous supratentorial intracerebral hemorrhage as an example

Evaluation metrics

Missing proportions

Machine learning methods

Traditional methods

LR

RF

NN

SVM

EL

Mode

KNN

MICE

Sensitivity

0.05

0.877

0.866

0.866

0.866

0.889

0.851

0.866

0.869

 

0.10

0.858

0.874

0.874

0.874

0.881

0.862

0.874

0.868

 

0.15

0.889

0.877

0.904

0.874

0.897

0.858

0.866

0.870

 

0.20

0.885

0.874

0.877

0.866

0.889

0.843

0.866

0.877

 

0.30

0.866

0.897

0.920

0.881

0.923

0.739

0.866

0.876

 

0.50

0.862

0.943

0.973

0.900

0.969

0.693

0.739

0.789

 

Average

0.873

0.889

0.902

0.877

0.908

0.808

0.846

0.858

AUC

0.05

0.913

0.912

0.913

0.912

0.913

0.901

0.912

0.911

 

0.10

0.909

0.914

0.915

0.914

0.911

0.904

0.915

0.912

 

0.15

0.919

0.917

0.924

0.916

0.919

0.893

0.911

0.913

 

0.20

0.913

0.918

0.916

0.914

0.916

0.891

0.915

0.912

 

0.30

0.902

0.921

0.933

0.921

0.934

0.860

0.910

0.907

 

0.50

0.887

0.952

0.947

0.942

0.950

0.855

0.860

0.875

 

Average

0.907

0.922

0.925

0.920

0.924

0.884

0.904

0.905

Kappa

0.05

0.562

0.555

0.555

0.555

0.569

0.519

0.555

0.555

 

0.10

0.547

0.557

0.557

0.557

0.561

0.526

0.557

0.554

 

0.15

0.565

0.558

0.574

0.555

0.591

0.507

0.551

0.561

 

0.20

0.568

0.561

0.563

0.556

0.592

0.506

0.557

0.564

 

0.30

0.547

0.566

0.579

0.556

0.619

0.491

0.547

0.569

 

0.50

0.507

0.627

0.630

0.622

0.645

0.556

0.491

0.514

 

Average

0.549

0.571

0.576

0.567

0.596

0.518

0.543

0.553