Skip to main content

Table 4 Performance comparison of classification techniques

From: Application of genetic algorithms and constructive neural networks for the analysis of microarray cancer data

 

GA

SFS

 

Classifier

Parameters

mean ± std

#genes

mean ± std

#genes

Leukemia

LDA

-

99.959 ± 0.07

12

97.609 ± 2.86

2

 

SVM

{polynomial,15,1,0.6,0}

99.982 ± 0.06

8

99.918 ± 0.52

4

 

NaiveBayes

{1,0}

99.974 ± 0.03

12

98.060 ± 2.19

3

 

C-MANTEC

{1000,0.01,4.5}

99.038 ± 0.25

7

98.837 ± 2.46

3

 

kNN

{1,Euclidean}

99.994 ± 0.02

10

99.844 ± 0.77

5

 

MLP

{3,0.5,50}

99.944 ± 0.05

5

95.784 ± 3.38

2

Lung

LDA

-

99.971 ± 0.03

5

99.057 ± 1.00

3

 

SVM

{linear,10,-,-,-}

100 ± 0

11

99.828 ± 0.70

3

 

NaiveBayes

{1,0}

99.998 ± 0.01

4

99.991 ± 0.07

3

 

C-MANTEC

{100000,0.25,2}

99.678 ± 0.08

6

99.673 ± 0.94

2

 

kNN

{1,Euclidean}

99.969 ± 0.02

4

99.969 ± 0.22

4

 

MLP

{4,0.1,50}

99.996 ± 0.01

4

99.778 ± 0.79

2

Colon

LDA

-

98.676 ± 0.35

11

87.179 ± 6.15

2

 

SVM

{polynomial,1,1,0.4,2}

89.917 ± 1.26

20

91.738 ± 5.21

5

 

NaiveBayes

{0,1}

90.583 ± 0.49

15

89.076 ± 7.79

4

 

C-MANTEC

{10000,0.01,1}

94.315 ± 0.48

11

87.593 ± 6.69

2

 

kNN

{3,cosine-similarity}

95.060 ± 0.38

19

93.577 ± 4.43

6

 

MLP

{5,0.3,50}

99.026 ± 0.30

12

88.733 ± 5.51

2

Breast

LDA

-

99.788 ± 0.12

15

74.169 ± 6.52

1

 

SVM

{polynomial,7,2,0.001,2}

99.744 ± 0.14

31

81.029 ± 5.80

3

 

NaiveBayes

{0,0}

97.759 ± 0.23

27

73.499 ± 6.34

1

 

C-MANTEC

{10000,0.01,1.5}

97.342 ± 0.39

23

76.645 ± 6.53

1

 

kNN

{3,Euclidean}

97.485 ± 0.30

34

80.975 ± 6.37

2

 

MLP

{4,0.3,50}

99.828 ± 0.09

18

79.191 ± 6.43

2

Ovarian

LDA

-

99.980 ± 0.01

4

100 ± 0

3

 

SVM

{polynomial,9,1,0.2,0}

100 ± 0

4

99.978 ± 0.13

4

 

NaiveBayes

{1,0}

99.951 ± 0.03

5

99.980 ± 0.13

4

 

C-MANTEC

{1000,0.3,1.5}

99.844 ± 0.05

4

99.659 ± 0.75

3

 

kNN

{1,Euclidean}

99.984 ± 0.01

4

99.982 ± 0.11

3

 

MLP

{5,0.3,50}

99.999 ± 0

3

100 ± 0

3

Prostate

LDA

-

99.720 ± 0.12

9

95.677 ± 2.81

4

 

SVM

{polynomial,5,1,3,1}

99.428 ± 0.31

20

98.622 ± 1.79

5

 

NaiveBayes

{0,0}

98.817 ± 0.16

14

98.331 ± 2.13

7

 

C-MANTEC

{1000,0.25,4}

98.681 ± 0.24

8

95.351 ± 3.40

4

 

kNN

{3,cosine-similarity}

99.633 ± 0.11

20

97.146 ± 2.28

6

 

MLP

{3,0.5,50}

99.996 ± 0.02

12

96.921 ± 2.37

4

  1. Performance comparison among the two different feature selection frameworks used (GA and SFS) and the six classifiers analyzed (LDA, SVM, NaiveBayes, C-MANTEC, kNN and MLP) for each cancer microarray dataset. The results correspond to the best simulation for each dataset, showing the accuracy for method in the format of mean ± standard deviation and the number of selected genes.