Table 5.
Confusion matrices representing the mean and one standard deviation calculated over 10 iterations when the ANN parameters are randomly initialized and (a) the same training and test data sets are used in each iteration and (b) different training and test data sets are randomly constructed in each iteration.
| (a)
| |||||||||
|---|---|---|---|---|---|---|---|---|---|
|
classified |
|||||||||
| M1 | M2 | M3 | M4 | M5 | M6 | M7 | M8 | ||
| true | M1 | 100 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 |
| M2 | 0 ± 0 | 98.2 ± 3.8 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 1.8 ± 3.8 | |
| M3 | 0 ± 0 | 0 ± 0 | 99.7 ± 1.0 | 0.3 ± 1.0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | |
| M4 | 0 ± 0 | 0.6 ± 1.9 | 0 ± 0 | 95.8 ± 2.9 | 3.6 ± 2.4 | 0 ± 0 | 0 ± 0 | 0 ± 0 | |
| M5 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 4.8 ± 2.9 | 95.2 ± 2.9 | 0 ± 0 | 0 ± 0 | 0 ± 0 | |
| M6 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 2.4 ± 4.0 | 0.9 ± 2.0 | 96.7 ± 3.9 | 0 ± 0 | 0 ± 0 | |
| M7 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0.9 ± 1.5 | 0.3 ± 1.0 | 0 ± 0 | 98.8 ± 1.6 | 0 ± 0 | |
| M8 | 0 ± 0 | 3.3 ± 4.2 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 96.7 ± 4.2 | |
| (b)
| |||||||||
|---|---|---|---|---|---|---|---|---|---|
|
classified |
|||||||||
| M1 | M2 | M3 | M4 | M5 | M6 | M7 | M8 | ||
| true | M1 | 100 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 |
| M2 | 0 ± 0 | 97.9 ± 1.5 | 0 ± 0 | 0.3 ± 1.0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 1.8 ± 1.6 | |
| M3 | 0 ± 0 | 0 ± 0 | 99.7 ± 1.0 | 0.3 ± 1.0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | |
| M4 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 92.0 ± 6.8 | 7.7 ± 6.8 | 0 ± 0 | 0.3 ± 1.0 | 0 ± 0 | |
| M5 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 4.3 ± 4.8 | 95.7 ± 4.8 | 0 ± 0 | 0 ± 0 | 0 ± 0 | |
| M6 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 1.2 ± 2.6 | 0 ± 0 | 98.8 ± 2.6 | 0 ± 0 | 0 ± 0 | |
| M7 | 0 ± 0 | 0 ± 0 | 0.3 ± 1.0 | 1.6 ± 4.9 | 1.5 ± 1.6 | 0 ± 0 | 96.6 ± 4.5 | 0 ± 0 | |
| M8 | 0 ± 0 | 5.2 ± 6.6 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 0 ± 0 | 94.8 ± 6.6 | |