Table 2.
10-fold cross-validated accuracies for the classification between HC and sdMCI. The table shows the mean accuracies (specificities; sensitivities) in percentages for all five frequency bands, the three regularization methods (GL, GLa, GLd) and the four classifiers evaluated. Accuracies higher than 80% are highlighted.
| HC vs sdMCI | ||||
|---|---|---|---|---|
| GL | GLa | GLd | ||
| alpha | Knn | 72.55 (79.31; 63.64) | 72.55 (68.97; 77.27) | 72.55 (72.41; 72.73) |
| LDA | 80.39 (93.10; 63.64) | 84.31 (89.66; 77.27) | 82.35 (93.10; 68.18) | |
| SVM | 64.71 (75.86; 50.00) | 76.47 (82.76; 68.18) | 76.47 (86.21; 63.64) | |
| SVMrbf | 82.35 (89.66; 72.73) | 80.39 (86.21; 72.73) | 80.39 (93.10; 63.64) | |
| Low beta | Knn | 72.55 (62.07; 86.36) | 72.55 (65.52; 81.82) | 68.63 (51.72; 90.91) |
| LDA | 70.59 (68.97; 72.73) | 74.51 (75.86; 72.73) | 72.55 (89.66; 50.00) | |
| SVM | 66.67 (65.52; 68.18) | 68.63 (75.86; 59.09) | 66.67 (68.97; 63.64) | |
| SVMrbf | 68.63 (72.41; 63.64) | 72.55 (79.31; 63.64) | 68.63 (72.41; 63.64) | |
| High beta | Knn | 70.59 (65.52; 77.27) | 72.55 (79.31; 63.64) | 74.51 (79.31; 68.18) |
| LDA | 82.35 (86.21; 77.27) | 76.47 (79.31; 72.73) | 76.47 (86.21; 63.64) | |
| SVM | 66.67 (75.86; 54.55) | 70.59 (75.86; 63.64) | 80.39 (86.21; 72.73) | |
| SVMrbf | 68.63 (93.10; 36.36) | 76.47 (96.55; 50.00) | 74.51 (86.21; 59.09) | |
| Full beta | Knn | 68.63 (62.07; 77.27) | 70.59 (68.97; 72.73) | 68.63 (62.07; 77.27) |
| LDA | 74.51 (82.76; 63.64) | 68.63 (75.86; 59.09) | 76.47 (82.76; 68.18) | |
| SVM | 72.55 (79.31; 63.64) | 64.71 (65.52; 63.64) | 68.63 (75.86; 59.09) | |
| SVMrbf | 76.47 (86.21; 63.64) | 66.67 (96.55; 27.27) | 72.55 (89.66; 50.00) | |
| Broadband | Knn | 74.51 (75.86; 72.73) | 72.55 (79.31; 63.64) | 80.39 (72.41; 90.91) |
| LDA | 78.43 (82.76; 72.73) | 74.51 (82.76; 63.64) | 86.27 (89.66; 81.82) | |
| SVM | 74.51 (75.86; 72.73) | 74.51 (75.86; 72.73) | 74.51 (82.76; 63.64) | |
| SVMrbf | 80.39 (82.76; 77.27) | 72.55 (79.31; 63.64) | 82.35 (89.66; 72.73) | |