Skip to main content

Table 4 Robustness for the EBD and FI discretization methods

From: Application of an efficient Bayesian discretization method to biomedical data

Classifier

C4.5

NB

Dataset

EBD (SEM)

FI (SEM)

EBD (SEM)

FI (SEM)

1

100.00% (0.00)

100.00% (0.00)

94.94% (0.89)

94.94% (0.97)

2

90.64% (0.77)

87.69% (0.86)

94.36% (0.98)

95.17% (1.05)

3

70.78% (2.00)

53.57% (2.10)

82.44% (1.10)

81.69% (1.14)

4

84.18% (0.76)

85.87% (0.77)

90.10% (1.09)

75.91% (1.09)

5

49.83% (2.01)

53.08% (2.18)

69.97% (1.20)

86.88% (1.12)

6

83.58% (1.34)

80.58% (1.42)

97.76% (1.12)

95.89% (0.92)

7

92.50% (1.18)

92.50% (1.18)

96.67% (0.86)

97.27% (0.86)

8

55.50% (2.16)

55.11% (2.06)

70.94% (1.48)

71.67% (1.43)

9

90.61% (0.95)

87.16% (0.99)

98.98% (0.74)

96.08% (0.94)

10

75.10% (1.48)

68.65% (1.39)

74.35% (2.05)

76.93% (1.81)

11

70.36% (0.95)

70.47% (0.93)

78.25% (1.22)

82.52% (1.20)

12

57.82% (2.22)

61.04% (2.21)

63.47% (1.87)

65.94% (1.88)

13

66.12% (0.39)

66.96% (0.37)

64.89% (1.05)

50.83% (1.02)

14

57.47% (0.94)

64.13% (1.06)

67.01% (1.08)

69.18% (1.08)

15

54.94% (0.72)

54.20% (0.74)

54.16% (1.75)

61.60% (1.70)

16

73.17% (1.66)

77.17% (1.79)

92.57% (1.39)

84.11% (1.38)

17

82.71% (1.35)

87.43% (1.21)

88.25% (1.56)

85.49% (1.60)

18

79.38% (0.57)

82.65% (0.57)

88.91% (0.72)

91.81% (0.83)

19

73.00% (1.48)

79.00% (1.30)

85.55% (1.31)

85.89% (1.29)

20

58.75% (2.09)

58.75% (2.08)

100.00% (0.00)

100.00% (0.00)

21

55.18% (1.26)

62.23% (1.13)

77.01% (0.60)

76.10% (0.57)

22

72.53% (0.96)

66.84% (1.03)

90.87% (0.89)

81.15% (0.93)

23

78.16% (1.04)

76.07% (0.99)

77.79% (1.67)

52.49% (1.73)

24

75.00% (1.78)

70.00% (1.75)

80.33% (1.64)

99.86% (1.70)

Average

72.55% (2.81)

72.81% (2.76)

81.72% (2.92)

82.40% (2.59)

  1. The mean and the standard error of the mean (SEM) for robustness for each dataset is obtained by 10 × 10 cross-validation. For each dataset, the higher robustness value is shown in bold font and equal robustness values are underlined.