Original title: Forecast U.S. Covid-19 Numbers by Open SIR Model with Testing
Authors: Bo Deng
In this article, the U.S. Covid-19 data shows frequent ups and downs within larger waves, but existing models don’t explain this pattern. The researchers introduce a modified SIR model to tackle these spikes and accurately predict future cases. Over 590 days, their model impressively fits the daily case numbers with only a 0.5 percent error. However, when predicting 28 days ahead, it underestimates the total cases by 20 percent due to a consistent error. By adjusting the model to address this issue, they manage to keep their 28-day forecasts within a 10 percent margin of error compared to the actual data. This work suggests improvements to better understand and forecast Covid-19 outbreaks in the U.S.
Original article: https://arxiv.org/abs/2311.10762