Table 2
Average accuracy, sensitivity and specificity, as well as their respective standard deviations (std, stdsens and stdspec) of the SVM five-fold classification along the ten iterations
| C | γ | Acc. (%) | Sens. (%) | Spec. (%) | std | stdsens | stdspec | |
|---|---|---|---|---|---|---|---|---|
| WSF | 500 | 10–4 | 73.47 | 78.71 | 68.14 | 0.90 | 1.13 | 1.26 |
| WCF4 | 50 | 10–4 | 73.66 | 77.15 | 70.12 | 1.25 | 1.38 | 1.68 |
| WCF13 | 250 | 10–4 | 75.95 | 77.86 | 73.96 | 0.87 | 1.63 | 1.04 |
| LBP1,8 | 50 | 10–3 | 68.34 | 70.49 | 66.21 | 2.54 | 2.18 | 3.51 |
 | 50 | 10–4 | 70.02 | 75.95 | 64.01 | 1.02 | 1.22 | 1.49 |
 | 10 | 0.01 | 74.22 | 81.97 | 66.37 | 0.70 | 1.39 | 1.57 |
 | 250 | 10–4 | 78.84 | 79.84 | 77.80 | 1.12 | 1.60 | 1.25 |
 | 100 | 10–4 | 78.13 | 78.62 | 77.58 | 1.16 | 2.07 | 1.55 |
| BoW | 5 | 10–4 | 81.16 | 79.31 | 82.97 | 1.72 | 2.20 | 2.57 |
| C | γ | Acc. (%) | Sens. (%) | Spec. (%) | std | stdsens | stdspec | |
|---|---|---|---|---|---|---|---|---|
| WSF | 500 | 10–4 | 73.47 | 78.71 | 68.14 | 0.90 | 1.13 | 1.26 |
| WCF4 | 50 | 10–4 | 73.66 | 77.15 | 70.12 | 1.25 | 1.38 | 1.68 |
| WCF13 | 250 | 10–4 | 75.95 | 77.86 | 73.96 | 0.87 | 1.63 | 1.04 |
| LBP1,8 | 50 | 10–3 | 68.34 | 70.49 | 66.21 | 2.54 | 2.18 | 3.51 |
| | 50 | 10–4 | 70.02 | 75.95 | 64.01 | 1.02 | 1.22 | 1.49 |
| | 10 | 0.01 | 74.22 | 81.97 | 66.37 | 0.70 | 1.39 | 1.57 |
| | 250 | 10–4 | 78.84 | 79.84 | 77.80 | 1.12 | 1.60 | 1.25 |
| | 100 | 10–4 | 78.13 | 78.62 | 77.58 | 1.16 | 2.07 | 1.55 |
| BoW | 5 | 10–4 | 81.16 | 79.31 | 82.97 | 1.72 | 2.20 | 2.57 |
The cost parameter for the SVM scheme (C) and the regularisation parameter of the radial basis function kernel (γ) for which the best results with each set of descriptors (i.e. first column on the left) were obtained, are provided. Abbreviations: acc., average accuracy; sens., sensitivity; spec., specificity.