Skip to main content
. 2022 Nov 11;11(11):1654. doi: 10.3390/biology11111654

Table 8.

Performance evaluation of different feature selection methods on RF-XGB.

Feature Selection Configuration Feature Number Test ACC (%) MCC (%) F1 (%) AUC (%)
All features 16 features 16 96.03 97.06 96.03 96.50
Univariate Feature 14 features 14 96.58 96.39 96.58 97.54
Univariate Feature 12 features 12 97.35 95.55 97.31 98.21
Correlation Matrix 0.01 threshold 15 96.70 96.56 96.69 97.58
Correlation Matrix 0.015 threshold 14 97.25 96.64 97.25 97.62
Correlation Matrix 0.025 threshold 12 97.35 95.55 97.31 98.21
PAC - 15 96.70 96.56 96.69 97.58
PAC - 14 97.25 96.64 97.25 97.62
PAC - 10 96.74 95.13 96.74 98.03
Wrapper Method 0.05 thresh 14 97.92 96.89 97.92 98.73
Wrapper Method 0.01 thresh 13 97.13 95.72 97.13 98.23
Wrapper Method 0.045 thresh 9 96.74 95.15 96.73 97.97
RF Threshold 0.25 14 96.90 96.76 96.90 97.48
RF Threshold 0.045 12 97.81 96.73 97.81 98.67
RF Threshold 0.05 10 98.05 97.27 98.05 98.91