Table 3.
Sensitivity and specificity of prediction samples for classifying NILM vs. SIL by PCA-SVM and GA-SVM based models.
| Algorithm | Sensitivity (%) | Specificity (%) |
|---|---|---|
| PCA-SVM-L | 60.0 | 33.3 |
| PCA-SVM-Q | 80.0 | 50.0 |
| PCA-SVM-P | 80.0 | 83.3 |
| PCA-SVM-RBF | 80.0 | 83.3 |
| PCA-SVM-MLP | 20.0 | 16.7 |
| GA-SVM-L | 80.0 | 50.0 |
| GA-SVM-Q | 80.0 | 16.7 |
| GA-SVM-P | 40.0 | 66.7 |
| GA-SVM-RBF | 40.0 | 66.7 |
| GA-SVM-MLP | 60.0 | 33.3 |
| SVM-RBF | 0 | 100 |
Five different kernels were applied: linear (L), quadratic (Q), 3rd order polynomial (P), radial basis function (RBF) and multilayer perceptron (MLP).