Skip to main content

Table 6 LOOCV accuracy and the number of genes used in classifiers (in parenthesis) for binary class expression datasets

From: TSG: a new algorithm for binary and multi-class cancer classification and informative genes selection

Method

Colon

Leuk

CNS

DLBCL

Lung

Pros1

Pros2

Pros3

GCM

Aver

TSG†

93.55

(2)

98.61

(2)

97.06

(2)

98.7

(2)

100

(2)

95.1

(2)

86.36

(10)

100

(2)

87.5

(7)

95.21

TSP*

91.10

(2)

93.80

(2)

77.90

(2)

98.10

(2)

98.30

(2)

95.10

(2)

67.60

(2)

97.00

(2)

75.40

(2)

88.26

k-TSP*

90.30

(2)

95.83

(18)

97.10**

(10)

97.40

(2)

98.90

(10)

91.18

(2)

75.00

(18)

97.00

(2)

85.40

(10)

92.01

DT*

77.42

(3)

73.61

(2)

67.65

(2)

80.52

(3)

96.13

(3)

87.25

(4)

64.77

(4)

84.85

(1)

77.86

(14)

78.90

NB*‡

56.45

100

82.35

80.52

97.79

62.75

73.86

90.91

84.29

80.99

k-NN*‡

74.19

84.72

82.35

89.61

98.34

74.51

73.86

93.94

86.79

84.26

SVM*‡

82.26

98.61

82.35

97.40

99.45

91.18

76.14

100

93.21

91.18

PAM*

89.52

(15)

94.03

(2296)

82.35

(4)

85.45

(17)

97.90

(9)

90.89

(47)

81.25

(13)

94.24

(701)

82.32

(47)

88.66

  1. *Results reported in Tan et al. [3]†Results obtained with our method (TSG) ‡NB, k-NN, SVM used entire set of genes
  2. **The 97.10 reported in Tan et al. [3] may be a result of rounding 97.06, which is the accuracy of correctly classifying 33 of the 34 samples.