Table 9 Comparison of PHONEME Indian results across different models.

From: Optimizing imbalanced learning with genetic algorithm

%

Metrics

SMOTE9

ADASYN11

GAN71

VAE63

SGA

EGA

SVMGA

20%

Accuracy

83.7

82.7

81.0

81.6

81.5

82.8

83.2

Precision

68.9

66.4

62.2

64.2

62.9

66.7

68.1

Recall

79.4

80.7

86.9

81.8

87.1

80.9

79.2

F1 Score

73.8

72.9

72.5

71.9

73.0

73.1

73.2

ROC AUC

82.4

82.1

82.8

81.7

83.1

82.3

82.4

40%

Accuracy

82.6

81.7

79.9

80.9

81.6

82.0

82.6

Precision

65.9

63.5

60.8

62.1

63.1

64.3

64.7

Recall

82.8

85.2

85.2

86.7

86.7

84.1

85.8

F1 Score

73.4

72.8

71.0

72.3

73.1

72.9

73.8

ROC AUC

82.7

82.7

81.5

82.6

83.1

82.6

83.5

60%

Accuracy

83.9

78.3

80.2

82.6

80.6

81.8

82.4

Precision

68.0

57.9

60.5

66.2

63.7

64.3

64.7

Recall

83.7

89.7

89.5

81.3

83.0

83.0

88.0

F1 Score

75.0

70.4

72.2

73.0

72.1

72.5

74.6

ROC AUC

83.8

81.7

82.9

82.2

83.0

82.2

84.3

80%

Accuracy

79.9

78.6

80.2

80.2

79.7

80.8

80.9

Precision

60.2

52.3

60.5

61.2

58.4

61.6

59.1

Recall

89.3

94.2

89.9

85.8

88.8

87.5

90.7

F1 Score

71.9

67.3

72.3

71.5

70.5

72.3

71.6

ROC AUC

82.6

80.1

83.1

81.9

82.5

82.7

83.8

100%

Accuracy

69.8

77.2

79.5

78.7

77.5

79.3

79.6

Precision

52.5

55.8

59.7

58.5

58.9

58.5

60.8

Recall

94.2

92.7

88.2

89.7

87.1

87.2

90.5

F1 Score

67.4

69.7

71.2

70.8

70.3

70.0

72.7

ROC AUC

76.9

81.7

82.1

82.0

80.3

82.3

82.7

  1. This table shows the performance metrics (accuracy, precision, recall, F1 score, and ROC AUC) for models trained with different synthetic data generation techniques (SMOTE, ADASYN, GAN, VAE, SGA, EGA, and SVMGA) on the PHONEME dataset at varying percentages of data usage.
  2. Significant values are in bold.