Table 11 Pairwise comparison of mean F1-scores across various different models on PHONEME Dataset.

From: Optimizing imbalanced learning with genetic algorithm

Model

SMOTE9

ADASYN11

GAN71

VAE63

SGA

EGA

SVMGA

SMOTE

72.30

\(+\) 1.68 (0.2710)

\(+\) 0.46 (0.7211)

\(+\) 0.40 (0.7078)

\(+\) 0.50 (0.6284)

\(+\) 0.14 (0.8744)

− 0.88 (0.4751)

ADASYN

− 1.68 (0.2710)

70.62

− 1.22 (0.3485)

− 1.28 (0.2558)

− 1.18 (0.1097)

− 1.54 (0.1769)

− 2.56 (0.0355)

GAN

− 0.46 (0.7211)

\(+\) 1.22 (0.3485)

71.84

− 0.06 (0.8925)

\(+\) 0.04 (0.9546)

− 0.32 (0.5565)

− 1.34 (0.0992)

VAE

− 0.40 (0.7078)

\(+\) 1.28 (0.2558)

\(+\) 0.06 (0.8925)

71.90

\(+\) 0.10 (0.8311)

− 0.26 (0.5383)

− 1.28 (0.0146)

SGA

− 0.50 (0.6284)

\(+\) 1.18 (0.1097)

− 0.04 (0.9546)

− 0.10 (0.8311)

71.80

− 0.36 (0.3974)

− 1.38 (0.0399)

EGA

− 0.14 (0.8744)

\(+\) 1.54 (0.1769)

\(+\) 0.32 (0.5565)

\(+\) 0.26 (0.5383)

\(+\) 0.36 (0.3974)

72.16

− 1.02 (0.1779)

SVMGA

\(+\) 0.88 (0.4751)

\(+\) 2.56 (0.0355)

\(+\) 1.34 (0.0992)

\(+\) 1.28 (0.0146)

\(+\) 1.38 (0.0399)

\(+\) 1.02 (0.1779)

73.18

  1. Diagonal entries show the average F1-score for each model. Off-diagonal entries represent the mean difference in F1-score between the row and column models, followed by the p-value from a paired t-test in parentheses.
  2. Significant values are in bold.
Back to article page