Table 5. MLP, RBF neural networks and linear SVM (), SVM with polynomial kernel () and SVM with RBF kernel () classifications using classical and non-linear/multi-resolution features of the HRV records from normal (N) and cardiovascular risk (R) subjects.
Features | Classifier | Se (%) | Sp (%) | Np (%) | Pp (%) | Ac (%) |
Statistical + Spectral | MLP | 66.67 | 60.00 | 64.29 | 62.50 | 63.33 |
RBFNN | 26.67 | 93.33 | 56.00 | 80.00 | 60.00 | |
SVM (Linear) | 72.73 | 86.36 | 76.00 | 84.21 | 79.55 | |
SVM (Polynomial kernel) | 68.18 | 70.45 | 68.89 | 69.77 | 69.32 | |
SVM (RBF kernel) | 68.18 | 74.24 | 70.00 | 72.58 | 71.21 | |
Non-linear + Multi-resolution | MLP | 80.00 | 100.00 | 83.33 | 100.00 | 90.00 |
RBFNN | 73.33 | 100.00 | 78.95 | 100.00 | 86.67 | |
SVM (Linear) | 95.45 | 77.27 | 94.44 | 80.77 | 86.36 | |
SVM (Polynomial kernel) | 88.64 | 81.82 | 87.80 | 82.98 | 85.23 | |
SVM (RBF kernel) | 90.91 | 83.33 | 90.16 | 84.51 | 87.12 |