Research Article
Air Pollution Concentration Forecast Method Based on the Deep Ensemble Neural Network
Table 4
The accuracy comparisons of different models with different optimized algorithms.
| Model | The number of neurons of dense layers | Algorithm | MAE | MAPE |
| NN | 15 | SGD | 7.13 | 22.38% | 15 | Adam | 7.11 | 19.89% | 15 | Nadam | 6.75 | 20.11% | RNN | 15 | SGD | 12.33 | 32.45% | 15 | Adam | 6.57 | 16.77% | 15 | Nadam | 6.58 | 17.29% | LSTM | 15 | SGD | 11.18 | 30.38% | 15 | Adam | 6.45 | 17.34% | 15 | Nadam | 6.67 | 17.47% | GRU | 15 | SGD | 9.19 | 25.00% | 15 | Adam | 6.56 | 19.84% | 15 | Nadam | 6.70 | 19.98% |
|
|