Table 5.
BPNN for BRPH age and amount classification based on K-fold cross-validation.
| Repeat Times | Nodes | LR | DF | MTT | Classification Accuracy | ||
|---|---|---|---|---|---|---|---|
|
| |||||||
| Training Set | Test Set | ||||||
| BPNN for BRPH age classification | 1 | 18 | 0.035 | 0.05 | 20,000 | 100% | 100% |
| 2 | 18 | 0.035 | 0.05 | 20,000 | 100% | 96.67% | |
| 3 | 18 | 0.034 | 0.05 | 20,000 | 100% | 96.67% | |
| 4 | 19 | 0.034 | 0.05 | 20,000 | 100% | 100% | |
| 5 | 19 | 0.036 | 0.05 | 20,000 | 100% | 90% | |
|
| |||||||
| Average accuracy | 100% | 96.67% | |||||
|
| |||||||
| BPNN for BRPH amount classification | 1 | 22 | 0.05 | 0.8 | 25,000 | 52.8% | 44% |
| 2 | 25 | 0.05 | 0.8 | 25,000 | 42.4% | 52% | |
| 3 | 25 | 0.05 | 0.8 | 25,000 | 48.8% | 52% | |
| 4 | 24 | 0.05 | 0.8 | 25,000 | 45.6% | 44% | |
| 5 | 25 | 0.05 | 0.8 | 25,000 | 47.2% | 48% | |
| 6 | 24 | 0.05 | 0.8 | 25,000 | 56.8% | 44% | |
|
| |||||||
| Average accuracy | 48.93% | 47.33% | |||||
Note: Nodes, the number of nodes in the hidden layer; LR, learning rate; DF, dynamic factor; MTT, the maximum number of training times.