Research Article
Influence of Training Set Selection in Artificial Neural Network-Based Propagation Path Loss Predictions
Table 3
Error statistics for the outdoor case with the ANN trained with set A and with set B.
| Test routes | Route 1 | Route 2 | Route 3 | Route 4 | Route 5 |
| Network trained with training Set A |
| Mean error | 10.03 | 8.09 | 11.41 | 5.76 | 3.86 | RMS error | 14.17 | 10.11 | 13.26 | 8.29 | 5.15 | std | 10.01 | 6.07 | 6.76 | 5.97 | 3.41 |
| Network trained with training Set B |
| Mean error | 6.66 | 6.70 | 5.40 | 4.26 | 1.77 | RMS error | 10.54 | 9.49 | 7.62 | 6.58 | 2.52 | std | 8.17 | 6.72 | 5.24 | 5.01 | 1.79 |
|
|