Table 3

The comparisons in seven gene selection methods (gene number = 30).

Feature Selection Method

Classifier

ALL

ARR

LYM

HBC

NCI60

MLL

GCM


No feature sel

SVM

91.94%

51.04%

95.16%

77.27%

63.33%

97.22%

51.52%

Naive Bayes

85.23%

49.57%

95.04%

70.11%

45.22%

93.13%

40.33%

mRMR-ReliefF

SVM

96.77%

81.43%

100%

95.45%

68.33%

98.61%

64.65%

Naive Bayes

95.97%

79.05%

100%

95.45%

61.67%

98.61%

61.11%

Maxrel

SVM

89.11%

74.53%

100%

72.73%

51.67%

77.78%

60.61%

Naive Bayes

88.71%

73.49%

100%

63.64%

48.33%

80.56%

46.97%

Information Gain

SVM

97.58%

80.13%

98.39%

100%

61.67%

98.67%

46.67%

Naive Bayes

92.74%

77.21%

93.55%

86.38%

60%

97.22%

47.47%

Sum Minority

SVM

93.95%

76.42%

98.39%

95.45%

55%

90.28%

55.05%

Naive Bayes

91.13%

74.32%

95.16%

81.82%

46.67%

91.67%

49.49%

Twoing Rule

SVM

96.77%

79.37%

98.39%

90.91%

61.67%

97.22%

45.96%

Naive Bayes

90.32%

72.19%

93.55%

86.36%

45%

95.83%

46.46%

F-statistic

SVM

97.17%

67.12%

96.77%

90.91%

63.33%

77.22%

39.10%

Naive Bayes

80.27%

71.55%

98.52%

85.41%

60.15%

80.13%

39.81%

GSNR

SVM

93.18%

77.24%

100%

95.45%

63.37%

90.25%

40.74%

Naive Bayes

90.11%

70.43%

100%

85.65%

58.25%

87.22%

39.81%


This table shows the classification results based on the 30 genes, which are selected from 7 different datasets using seven feature selection methods, named mRMR-ReliefF, Maxrel, information gain, sum minority, twoing rule, F-statistic, GSNR.

Zhang et al. BMC Genomics 2008 9(Suppl 2):S27   doi:10.1186/1471-2164-9-S2-S27

Open Data