Table 10 Comparative analysis of the proposed criminal suspect identification model against ML and DL models on the VGGFace2 dataset.

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

Model

Accuracy

F-measure

G-means

AUC

MISSL28

69.542 ± 0.005

74.202 ± 0.051

72.929 ± 0.014

0.671 ± 0.037

CNBA29

70.825 ± 0.005

75.562 ± 0.028

76.221 ± 0.027

0.682 ± 0.088

FAHP33

71.930 ± 0.048

76.827 ± 0.055

77.533 ± 0.014

0.707 ± 0.051

DL-ACO36

72.255 ± 0.042

74.913 ± 0.043

75.572 ± 0.051

0.711 ± 0.088

DCNN37

80.878 ± 0.051

81.940 ± 0.071

82.599 ± 0.016

0.740 ± 0.071

MTCNN38

82.485 ± 0.098

82.146 ± 0.098

82.842 ± 0.059

0.753 ± 0.025

DNVPT4

87.591 ± 0.098

86.377 ± 0.004

87.044 ± 0.092

0.811 ± 0.068

FECNN39

84.164 ± 0.095

83.056 ± 0.067

84.746 ± 0.020

0.767 ± 0.055

Wavelet2

86.126 ± 0.034

86.693 ± 0.046

86.984 ± 0.066

0.802 ± 0.023

CLSTM40

85.192 ± 0.053

83.113 ± 0.080

84.822 ± 0.035

0.772 ± 0.064

QN-FR41

80.824 ± 0.089

82.779 ± 0.018

83.450 ± 0.077

0.768 ± 0.062

DNN1

86.494 ± 0.016

84.514 ± 0.058

85.185 ± 0.031

0.782 ± 0.025

YOLOv8-FI42

82.120 ± 0.014

83.512 ± 0.093

84.247 ± 0.063

0.778 ± 0.078

FVG-FR43

84.631 ± 0.020

85.139 ± 0.051

86.836 ± 0.001

0.783 ± 0.081

QWE-DNN44

85.804 ± 0.021

86.552 ± 0.002

86.620 ± 0.036

0.816 ± 0.076

FacialCueNet45

81.436 ± 0.060

82.738 ± 0.078

83.434 ± 0.014

0.828 ± 0.012

GAN-DSAEAN47

86.451 ± 0.064

87.680 ± 0.021

88.405 ± 0.023

0.835 ± 0.094

Proposed

91.511 ± 0.005

92.202 ± 0.051

93.929 ± 0.014

0.872 ± 0.037

Back to article page