Table 3. Performance of prediction models for seasonal influenza outbreaks after 26 weeks in the United States from 2015 to 2019.
| R2 | RMSE (forecast values) | RMSE (Peak timing) | RMSE (Peak amplitude) | ||
|---|---|---|---|---|---|
| Influenza-like illness | Previous season | 0.487 | 1.1 | 5.5 | 2.2 |
| LR | 0.720 | 0.8 | 2.5 | 1.5 | |
| ARIMAX | 0.714 | 0.8 | 2.9 | 1.6 | |
| ANN | 0.758* | 0.7* | 1.8* | 0.8* | |
| Total Influenza viruses | Previous season | 0.346 | 4802.8 | 5.8 | 6657.8 |
| LR | 0.792 | 2707.6 | 1.7 | 5574.5* | |
| ARIMAX | 0.806* | 2618.2* | 1.7 | 6726.4 | |
| ANN | 0.738 | 3039.6 | 1.7* | 7544.3 | |
| Influenza A virus | Previous season | 0.289 | 4190.9 | 5.8 | 4901.0 |
| LR | 0.777 | 2347.8 | 1.9 | 4694.7 | |
| ARIMAX | 0.792 | 2265.2 | 1.6 | 5355.9 | |
| ANN | 0.798* | 2231.9* | 1.2* | 4545.3* | |
| Influenza B virus | Previous season | -0.238 | 1880.8 | 3.7 | 5142.3 |
| LR | 0.427* | 1279.4* | 3.0 | 1453.6* | |
| ARIMAX | 0.352 | 1360.4 | 2.7 | 3163.0 | |
| ANN | 0.403 | 1306.3 | 2.4* | 1512.0 |
R2, Coefficient of determination; RMSE, Root-mean-square error; LR, Linear regression; ARIMAX, Auto Regressive Integrated Moving Average; ANN, artificial neural network.
* Best value among previous season, LR, ARIMAX, and ANN.
Units for the RMSE (forecast values) are percentage of visits for influenza-like illness and number of total influenza, influenza A, and influenza B viruses.
Units for the RMSE (Peak timing) are week.
Units for the RMSE (peak amplitude) are percentage of visits for influenza-like illness and number of total influenza, influenza A, and influenza B viruses.