Skip to main content

Table 6 The F-score of the proposed cost-sensitive method compared to other methods

From: Improved cost-sensitive representation of data for solving the imbalanced big data classification problem

Dataset

F-score (%)

Baseline SVM

S-MVML-LA

GMEB

WJMI

IGIS

IWFS

Proposed

Iris

66.16

74.23 (4)

94.98 (2)

NA

88.30 (3)

NA

95.60 (1)

Cancer

94.50

96.28 (3)

95.19 (5)

100 (1)

93.96 (6)

100 (1)

96.02 (4)

Ionosphere

40.81

68.14 (4)

82.37 (2)

NA

74.53 (3)

NA

83.97 (1)

Wine

18.98

82.74 (4)

95.54 (2)

NA

93.57 (3)

NA

98.31 (1)

Pid

40.81

39.37 (4)

72.18 (2)

NA

71.7 (3)

NA

72.57 (1)

Wdbc

82.99

88.34 (4)

94.42 (3)

NA

95.51 (2)

NA

96.73 (1)

Musk

83.11

84.91 (4)

80.21 (5)

85.85 (3)

82.34 (6)

87.06 (2)

87.86 (1)

Dermatology-6

48.56

82.95 (6)

96.33 (2)

87.51 (4)

89.41 (3)

86.2 (5)

97.90 (1)

FuelCons

46.05

54.73 (6)

67.21 (4)

77.49 (1)

66.68 (5)

76.92 (2)

72.51 (3)

Movement_libras

33.58

41.08 (6)

79.76 (3)

67.99 (5)

81.80 (1)

80.71 (2)

76.23 (4)

Sonar

80.38

78.81 (4)

85.85 (1)

77.77 (5)

75.28 (6)

82.61 (3)

85.76 (2)

SPECTF

44.16

44.17 (5)

44.17 (5)

62.39 (3)

44.26 (4)

67.39 (2)

66.36 (1)

Mnist

91.01

76.94 (3)

90.25 (2)

70.28 (5)

NA

75.85 (4)

91.62 (1)

Colon

82.13

82.14 (4)

65.95 (5)

96.18 (2)

NA

98.1 (1)

82.72 (3)

Caltch101

52.13

48.59 (1)

45.26 (3)

22.27 (5)

NA

17.75 (4)

48.43 (2)

DLBCL77

77.27

42.78 (5)

80.48 (4)

100 (1)

NA

100 (1)

96.73 (3)

Kddcup-rootkit-imap-vs-back

49.75

69.21 (3)

98.87 (1)

NA

NA

NA

96.65 (2)

Kddcup-buffer-overflow-vs-back

49.66

79.13 (3)

100 (1)

NA

NA

NA

100 (1)

Kddcup-guess-passwd-vs-satan

99.06

85.12 (3)

98.27 (2)

NA

NA

NA

100 (1)

Kddcup-land-vs-satan

78.03

86.04 (3)

100 (1)

NA

NA

NA

100 (1)

Cleveland

14.01

13.93 (3)

23.89 (2)

NA

NA

NA

29.51 (1)

Average rank

 

3.88

2.78

3.27

3.66

2.68

1.76

  1. The numbers in bold are the best accuracies achieved on the dataset