Skip to main content
. 2022 Feb 23;12(3):199. doi: 10.3390/metabo12030199

Table 1.

Performance comparison of feature selection strategies under 10-time 10-fold cross-validation. The number of features for each set is given in parentheses. The best performance across each 10-fold CV is highlighted in bold.

Initial (1446) Variance (1074) Correlation (439) RF Importance (64) RFECV (57) Final (8)
1st 10-Fold 80.06 ± 1.49 80.25 ± 2.17 80.18 ± 1.67 80.34 ± 1.05 80.2 ± 0.89 80.94 ± 1.07
2nd 10-Fold 80.67 ± 2.08 80.26 ± 1.87 80.56 ± 2.08 80.2 ± 3.01 80.58 ± 2.7 81.23 ± 2.04
3rd 10-Fold 80.81 ± 1.17 80.64 ± 1.83 79.48 ± 0.96 80.54 ± 2.04 80.33 ± 2.34 81.02 ± 1.86
4th 10-Fold 81.19 ± 1.81 80.14 ± 1.67 79.64 ± 1.51 80.35 ± 1.22 80.12 ± 1.52 80.52 ± 1.7
5th 10-Fold 80.66 ± 1.18 81.17 ± 1.23 79.55 ± 1.33 81.29 ± 1.2 81.13 ± 1.28 80.98 ± 1.59
6th 10-Fold 80.41 ± 1.12 80.13 ± 1.22 80.82 ± 1.59 81.15 ± 1.3 80.64 ± 1.05 81.36 ± 1.46
7th 10-Fold 81.42 ± 0.58 80.92 ± 1.61 80.67 ± 1.73 80.12 ± 1.55 80.85 ± 1.59 81.18 ± 1.19
8th 10-Fold 80.63 ± 1.53 81.02 ± 1.15 80.95 ± 0.87 80.58 ± 1.61 80.61 ± 2.34 81.01 ± 1.6
9th 10-Fold 81.13 ± 1.28 80.25 ± 1.43 80.43 ± 1.44 79.86 ± 1.11 79.76 ± 1.1 81.04 ± 0.96
10th 10-Fold 80.54 ± 1.08 80.24 ± 1.74 80.52 ± 0.92 80.68 ± 1.29 80.68 ± 1.32 81.03 ± 1.12