Skip to main content
. 2021 Feb 5;15:605115. doi: 10.3389/fnins.2021.605115

TABLE 4.

10-fold cross-validation binary mean classification performance for HC against MCI using RELM classifier using different feature selection methods.

Feature selection method Performance metrics Mean (%) Standard deviation Max (%) Min (%)
LASSO Accuracy 96.11 0.859 96.88 91.33
Sensitivity 95.03 1.080 95.93 89.84
Specificity 97.18 0.798 97.84 92.93
F-measure 0.973
FSASL Accuracy 85.85 0.9129 87.503 80.88
Sensitivity 79.27 0.986 81.484 76.01
Specificity 92.03 1.4433 93.308 85.40
F-measure 0.937
LLCFS Accuracy 82.29 0.624 83.39 80.77
Sensitivity 77.54 0.73 78.566 75.408
Specificity 86.81 1.081 88.41 83.74
F-measure 0.85
CFS Accuracy 80.67 1.68 90.427 74.48
Sensitivity 88.43 2.328 74.968 80.49
Specificity 72.38 1.3677 74.486 68.58
F-measure 0.795
SVM-RFE Accuracy 80.20 0.920 82.001 77.89
Sensitivity 84.00 1.207 85.99 79.86
Specificity 75.91 1.251 77.806 73.29
F-measure 0.815