Table 2.
Healthy vs. impaired classification results of individual tasks.
| Task: Features | Attribute Selection | Accuracy (10-Fold Cross-Validation) |
Speed (s) |
|---|---|---|---|
| T1: 9 | PCA (VC = 0.85) | J48:84.90%, SVM (RBF): 84.90% |
Instant |
| T2: 10 | WSE (VC = 0.60) | J48 (NBM): 81.13% | 0.07 |
| T3: 28 | WSE | RF (KNN): 77.35% | 1.72 |
| T4 (spiral following): 22 | WSE | LR (LDA): 84.90%, | 1.49 |
| ANN (RF): 82.07% | 36.06 | ||
| T4 (spiral drawing): 22 | WSE | ANN (KNN): 87.73% | 2.16 |
| RF (FLDA): 86.79% | 1.24 | ||
| T9: 30 | PCA (VC = 0.75) CAE |
LMT: 91.50% | 0.04, |
| ANN: 90.56% | 0.07 | ||
| RF: 90.56% | 0.02 | ||
| T10: 24 | CAE WSE |
KNN: 90.56% | Instant |
| ANN (KNN): 89.62% | 2.36 | ||
| T11: 2 | CAE | RF: 74.52% | 0.02 |
| T12: 33 | CAE WSE |
RF: 83.09% | 0.03 |
| ANN (KNN): 82.07% | 5.84 | ||
| T13: 25 | WSE | J48 (FLDA): 83.96% | 0.50 |
| T15: 4 | CAE | SVM (RBF): 78.3% | Instant |
| Spelling (T7, T8, T11): 3 | WSE | LMT (RF): 74.52% | 2.76 |
| SAGE (T6, T7, T8, T9, T10, T11, T12, T13, T0): 10 | PCA (VC = 0.50) CAE |
LMT: 84.90% | 0.01 |
| SMO: 84.90% | 0.03 | ||
| Duration (all tests): 16 | WSE | FLDA (FLDA): 89.62% | 0.66 |
PCA (VC)—Principal component analysis with variance covered (VC); WSE—Wrapper subset evaluation; CAE—Correlation attribute evaluation; RF—Random forest; KNN—K-nearest Neighbor; SVM—Support vector machine; RBF—Radial basis function; LR—Logistic regression; NBM—Naive Bayes multinomial; LMT—Logistic model trees.