National and Regional Influenza-Like-Illness Forecasts for the USA

Michal Ben-Nun; Pete Riley; James Turtle; David P. Bacon; Steven Riley

. CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/309021 doi: bioRxiv preprint Forecasts based on the mechanistic model performed better than the historic NULL 379 model for the peak intensity (Fig 3 A /B/C/D). Two weeks before the peak week (and 380 three weeks early in the season) we started predicting the correct peak intensity of 5.1% 381 (to within ±0.5%). The mean and median historic values are significantly lower (4.4% 382 and 4.1% respectively) and outside the ±0.5% range. Our apparent forecast 383 performance for intensity appears to drop off at the end of the season. However, this is 384 an artifact of the forecasting work flow. Once the peak had clearly passed, the final 385 model was selected for reasons other than the peak intensity and the already-observed 386 peak intensity was submitted.