Table 2.
Ten-fold cross-validation performance.
| Method | Atlas | Feature | Number of subjects | ACC(SD)% | AUC(SD)% |
|---|---|---|---|---|---|
| GaussianNB | MODL | Tangent | 866 | 64.0 (5.7) | 65.0 (6.7) |
| RandomF | MODL | Tangent | 866 | 62.8 (3.3) | 64.9 (5.5) |
| Logistic-L2 | MODL | Tangent | 866 | 64.8 (3.6) | 69.0 (5.3) |
| ridge | MODL | Tangent | 866 | 67.9 (3.2) | 70.0 (5.2) |
| KNN | MODL | Tangent | 866 | 65.8 (6.2) | 66.6 (5.9) |
| SVC | MSDL | Tangent | 871 | 66.9 (2.7) | - (-) |
| LSTM | CC200 | Time Series | 1,112 | 68.5 (5.5) | - (-) |
| InvNet | CC200 | Correlation | 1,035 | 71.0 (-) | - (-) |
| DNN | CC200 | Correlation | 1,035 | 70.0 (-) | - (-) |
| AttentionET | CC200 | Correlation | 1,054 | 72.2 (-) | - (-) |
| GCN | HO | Correlation | 871 | 70.4 (3.9) | 75.0 (4.6) |
| FGDN | HO | Correlation | 866 | 67.9 (5.5) | 69.5 (6.4) |
| FGDN | HO | Tangent | 866 | 71.4 (4.3) | 77.7 (5.3) |
| FGDN | AAL | Correlation | 866 | 65.9 (3.5) | 71.8 (5.1) |
| FGDN | AAL | Tangent | 866 | 70.8 (5.2) | 79.4 (5.0) |
| FGDN | MODL | Correlation | 866 | 68.0 (3.4) | 69.8 (5.2) |
| FGDN | MODL | Tangent | 866 | 71.8 (4.5) | 76.3 (5.6) |
The bold values indicate the largest values.