Table 16 Comparative analysis of the proposed DE against other basic and metaheuristic optimization algorithms on the CelebA dataset.

From: Reinforcement learning-driven feature selection enhanced by an evolutionary approach tuning for criminal suspect identification

Model

Accuracy

F-measure

G-means

AUC

Random search

67.511 ± 0.052

69.202 ± 0.046

70.929 ± 0.018

0.658 ± 0.066

Grid search

69.263 ± 0.095

71.753 ± 0.024

72.458 ± 0.026

0.662 ± 0.025

Hyperband

74.050 ± 0.037

76.855 ± 0.057

77.567 ± 0.036

0.672 ± 0.046

BO

80.869 ± 0.099

81.703 ± 0.086

82.412 ± 0.074

0.783 ± 0.023

SSA

72.314 ± 0.015

77.739 ± 0.079

78.428 ± 0.069

0.709 ± 0.059

HMS

73.476 ± 0.055

78.699 ± 0.085

79.397 ± 0.029

0.720 ± 0.029

COA

75.264 ± 0.042

80.000 ± 0.001

80.663 ± 0.099

0.736 ± 0.071

FA

76.976 ± 0.095

81.403 ± 0.043

82.091 ± 0.023

0.750 ± 0.075

BA

78.433 ± 0.024

82.610 ± 0.034

83.300 ± 0.075

0.756 ± 0.060

ABC

79.126 ± 0.049

83.604 ± 0.003

84.290 ± 0.011

0.771 ± 0.088

DE

80.147 ± 0.089

84.885 ± 0.053

85.585 ± 0.001

0.782 ± 0.023

Proposed DE

87.951 ± 0.096

89.409 ± 0.087

90.193 ± 0.030

0.829 ± 0.073