Table 6.
Model classification accuracy with added features.
| Class | Feature transformation | Classifier | Training accuracy | Testing accuracy | Sensitivity (training) | Specificity (training) | Sensitivity (testing) | Specificity (testing) |
|---|---|---|---|---|---|---|---|---|
| 1 (cutoff = KSS 6) | Principal Component Analysis | Logistics regression | 71.90% | 78.00% | 76.90% | 65.90% | 76.20% | 80.00% |
| SVM (Linear) | 74.00% | 87.80% | 75.00% | 72.70% | 81.00% | 95.00% | ||
| SVM (Quadratic) | 74.00% | 87.80% | 82.70% | 63.60% | 85.70% | 90.00% | ||
| SVM (Cubic) | 68.80% | 82.90% | 80.80% | 54.50% | 81.00% | 85.00% | ||
| SVM (Fine gaussian) | 77.10% | 82.90% | 73.10% | 81.80% | 85.70% | 80.00% | ||
| SVM (Medium gaussian) | 76.00% | 87.80% | 78.80% | 72.70% | 85.70% | 90.00% | ||
| SVM (Coarse gaussian) | 61.50% | 78.00% | 92.30% | 25.00% | 95.20% | 60.00% | ||
| No Feature Transformation | Logistics regression | 70.80% | 82.90% | 73.10% | 68.20% | 85.70% | 80.00% | |
| SVM (Linear) | 77.10% | 90.20% | 73.10% | 81.80% | 90.50% | 90.00% | ||
| SVM (Quadratic) | 72.90% | 80.50% | 76.90% | 68.20% | 85.70% | 75.00% | ||
| SVM (Cubic) | 65.60% | 68.30% | 65.40% | 65.90% | 71.40% | 65.00% | ||
| SVM (Fine gaussian) | 61.50% | 70.70% | 71.20% | 50.00% | 85.70% | 55.00% | ||
| SVM (Medium gaussian) | 65.60% | 85.40% | 69.20% | 61.40% | 90.50% | 80.00% | ||
| SVM (Coarse gaussian) | 63.50% | 75.60% | 88.50% | 34.10% | 95.20% | 55.00% | ||
| 2 (cutoff = KSS 7) | Principal Component Analysis | Logistics regression | 80.20% | 80.50% | 85.90% | 64.00% | 96.70% | 36.40% |
| SVM (Linear) | 83.30% | 80.50% | 93.00% | 56.00% | 96.70% | 36.40% | ||
| SVM (Quadratic) | 79.20% | 80.50% | 87.30% | 56.00% | 96.70% | 36.40% | ||
| SVM (Cubic) | 78.10% | 78.00% | 87.30% | 52.00% | 96.70% | 27.30% | ||
| SVM (Fine gaussian) | 77.10% | 78.00% | 90.10% | 40.00% | 96.70% | 27.30% | ||
| SVM (Medium gaussian) | 82.30% | 80.50% | 93.00% | 52.00% | 96.70% | 36.40% | ||
| SVM (Coarse gaussian) | 74.00% | 73.20% | 100.00% | 0.00% | 100.00% | 0.00% | ||
| No Feature Transformation | Logistics regression | 78.10% | 78.00% | 84.50% | 60.00% | 90.00% | 45.50% | |
| SVM (Linear) | 82.30% | 85.40% | 91.50% | 56.00% | 100.00% | 45.50% | ||
| SVM (Quadratic) | 80.20% | 80.50% | 85.90% | 64.00% | 90.00% | 54.50% | ||
| SVM (Cubic) | 76.00% | 75.60% | 81.70% | 60.00% | 83.30% | 54.50% | ||
| SVM (Fine gaussian) | 76.00% | 73.20% | 97.20% | 16.00% | 100.00% | 0.00% | ||
| SVM (Medium gaussian) | 80.20% | 75.60% | 94.40% | 40.00% | 96.70% | 18.20% | ||
| SVM (Coarse gaussian) | 74.00% | 73.20% | 100.00% | 0.00% | 100.00% | 0.00% |