Skip to main content

Table 3 The bootstrap based classification accuracies, stability index and number of reselected genes in all bootstrap samples of the SVM-classifier and the LDA-classifier based on all tested gene selection methods, on the SRBCT dataset

From: Stable feature selection and classification algorithms for multiclass microarray data

Selection method

stability index

Reselected genes

Classification method

Best result

      

SVM OvO

SVM OvR

MSVM

LDA

   
   

s 1

s 2

 

acc

acc

acc

acc

acc

accL

accH

SIMPLS

OvO

1 comp.

0.71

0.60

127

0.867

0.977

0.976

0.941

0.977

0.913

1.000

 

OvR

1 comp.

0.74

0.59

137

0.882

0.980

0.980

0.952

0.980

0.912

1.000

 

MCLASS

1 comp.

0.58

0.39

240

0.807

0.930

0.932

0.829

0.932

0.797

1.000

 

OvO

5 comp.

0.52

0.36

204

0.831

0.970

0.968

0.929

0.970

0.896

1.000

 

OvR

5 comp.

0.63

0.43

173

0.867

0.981

0.979

0.964

0.981

0.926

1.000

 

MCLASS

5 comp.

0.72

0.58

133

0.856

0.979

0.978

0.945

0.979

0.903

1.000

 

OvO

10 comp.

0.40

0.25

239

0.817

0.961

0.958

0.910

0.961

0.873

1.000

 

OvR

10 comp.

0.45

0.23

237

0.839

0.977

0.974

0.942

0.977

0.912

1.000

 

MCLASS

10 comp.

0.63

0.41

133

0.832

0.975

0.973

0.931

0.975

0.912

1.000

NIPALS

OvO

1 comp.

0.71

0.60

127

0.867

0.977

0.976

0.941

0.977

0.913

1.000

 

OvR

1 comp.

0.74

0.59

137

0.882

0.980

0.980

0.952

0.980

0.912

1.000

 

MCLASS

1 comp.

0.58

0.39

240

0.807

0.930

0.932

0.829

0.932

0.797

1.000

 

OvO

5 comp.

0.65

0.45

152

0.726

0.920

0.914

0.775

0.920

0.785

1.000

 

OvR

5 comp.

0.68

0.47

129

0.748

0.935

0.932

0.787

0.935

0.819

1.000

 

MCLASS

5 comp.

0.69

0.48

133

0.805

0.965

0.962

0.861

0.965

0.873

1.000

 

OvO

10 comp.

0.66

0.43

137

0.711

0.910

0.905

0.745

0.910

0.756

1.000

 

OvR

10 comp.

0.65

0.48

119

0.687

0.886

0.885

0.687

0.886

0.726

0.980

 

MCLASS

10 comp.

0.67

0.46

117

0.723

0.912

0.909

0.746

0.912

0.776

1.000

RFE

OvO

 

0.57

0.36

262

0.925

0.983

0.982

0.972

0.983

0.919

1.000

 

OvR

 

0.40

0.32

338

0.923

0.982

0.981

0.963

0.982

0.926

1.000

T-TEST

OvO

 

0.29

0.22

438

0.925

0.971

0.969

0.964

0.971

0.899

1.000

 

OvR

 

0.47

0.37

498

0.929

0.972

0.971

0.973

0.973

0.899

1.000

GS

OvO

 

0.51

0.39

533

0.921

0.969

0.968

0.964

0.969

0.889

1.000

 

OvR

 

0.64

0.54

367

0.927

0.980

0.977

0.979

0.980

0.923

1.000

 

MCLASS

 

0.57

0.48

407

0.921

0.959

0.957

0.959

0.959

0.873

1.000

  1. aThe number of selected genes is set to 30. In last 3 columns is the best accuracy rate together with their bootstrap based standard deviations (accL,accH). The number of reselected genes is the sum of non zero bootstrap-based feature ranked genes (BBFR)