Table 3 Mean Accuracy, Sensitivity, Specificity and Area under ROC curve in the 3-CV sets for the ANN architectures obtained by the PDM-ANN and the GA-ANN (T = Inf and T = 30)
Measurement | ANN architecture | Mean Value in 3-CV Training Sets | Mean Value in 3-CV Testing Sets |
|---|---|---|---|
| Â | PDM-ANN (32 factors) | 95.56 | 60.22 |
Accuracy (%) | GA-ANN, T = Inf (32 factors) | 97.67 | 60.69 |
| Â | GA-ANN, T = 30 (25 factors) | 97.10 | 61.46 |
| Â | PDM-ANN (32 factors) | 98.14 | 69.15 |
Sensitivity (%) | GA-ANN, T = Inf (32 factors) | 99.39 | 70.79 |
| Â | GA-ANN, T = 30 (25 factors) | 98.90 | 69.80 |
| Â | PDM-ANN (32 factors) | 91.15 | 46.08 |
Specificity (%) | GA-ANN, T = Inf (32 factors) | 94.73 | 44.62 |
| Â | GA-ANN, T = 30 (25 factors) | 94.54 | 48.63 |
| Â | PDM-ANN (32 factors) | 0.941 | 0.580 |
Area under ROC curve | GA-ANN, T = Inf (32 factors) | 0.969 | 0.574 |
| Â | GA-ANN, T = 30 (25 factors) | 0.964 | 0.608 |