Skip to main content

Table 7 Performance measures of data mining algorithm at different levels of significance on A & C conditions

From: Comparative study of classification algorithms for immunosignaturing data

SIGNIFICANCE

p < 5 x 10-4

p < 5 x 10-3

p < 5 x 10-2

 

Algorithm

Acc.

Sp

Sn

AUC

Acc.

Sp

Sn

AUC

Acc.

Sp

Sn

AUC

Avg.

Naïve Bayes

91.3

91.7

91.0

0.94

96.0

100

90.9

0.99

91.3

100

81.8

0.95

93.5

VFI

95.6

100

90.0

0.97

95.6

100

90.0

0.97

87.0

83.3

90.0

0.95

93.4

MLP

86.9

91.7

81.8

0.97

95.6

100

90.9

0.98

dnf

dnf

dnf

dnf

92.7*

SVM

95.6

100

90.9

0.96

95.7

100

90.9

0.96

73.9

75.0

72.7

0.74

88.4

Hyper Pipes

95.7

100

90.9

0.99

82.6

91.7

72.7

0.90

78.2

83.3

72.7

0.83

86.6

Logistic R.

86.0

91.7

81.8

0.96

95.7

100

90.9

0.92

69.6

83.3

54.5

0.76

84.8

KNN

91.3

100

81.8

0.92

91.3

100

81.8

0.94

65.2

66.7

63.6

0.72

83.3

Bayes Net

95.7

100

90.9

0.99

82.6

83.3

81.8

0.92

69.6

66.7

72.7

0.64

83.2

Random Forest

87.0

83.3

90.9

0.93

82.6

83.3

81.8

0.91

69.5

66.7

72.7

0.75

81.4

K means

69.6

83.3

54.5

0.69

95.7

100

90.9

0.95

60.9

63.6

63.6

0.63

75.7

M5P

91.3

91.7

90.9

0.86

65.2

58.3

72.7

0.72

65.2

58.3

72.7

0.56

73.4

LDA

91.3

100

81.8

0.97

65.2

71.7

58.6

0.77

17.4

25.0

100

0.52

69.7

K star

73.9

91.7

54.5

0.93

78.2

100

54.5

0.82

47.8

0.0

100

0.50

68.8

SLR

87.0

83.3

90.9

0.89

73.9

75.0

72.7

0.74

43.5

41.7

45.5

0.45

68.5

J48

69.6

66.7

72.7

0.76

69.6

58.3

81.8

0.77

60.9

58.3

63.6

0.66

68.4

ASC

65.6

66.7

72.7

0.76

69.6

66.7

72.7

0.76

47.8

66.7

27.3

0.49

63.1

Random Tree

73.9

91.7

54.5

0.73

73.9

66.7

81.8

0.74

34.8

33.3

36.4

0.35

60.8

  1. Acc: Accuracy, Sp: Specificity, Sn: Sensitivity, AUC: Area under ROC curve, Avg: Average score in % for each algorithms, dnf: Did not Finish”, * denotes Avg. from 3 significance levels. Measures >90% are marked in bold.