Table 7 Performance Comparison of Feature Selection Methods on Raspberry Pi 4 (Edge Deployment).

From: Integrating feature selection and explainable CNN for identification and classification of pests and beneficial insects

Method

Accuracy (%)

F1-score

Latency (s)

Feature Reduction (%)

No Feature Selection

88.4

0.87

2.35

0

Chi2 Filter

90.1

0.89

1.78

45

ReliefF

91.0

0.90

1.95

48

RFE (Wrapper)

91.7

0.91

2.20

52

LASSO (Embedded)

90.5

0.90

1.83

50

Hybrid (SHAP+PFI)

94.2

0.93

1.62

63

Back to article page