ReCOVER: Accurate Predictions and Scenario Projections for COVID-19 Epidemic Response

Created by

Ajitesh Srivastava ([email protected])

Code contributors

Ajitesh Srivastava, Frost Xu, Bob Xiaochen Yang, Jamin Chen

Approach

We use our own epidemic model called SI-kJalpha, preliminary version of which we have successfully used during DARPA Grand Challenge 2014. Our forecast appears on the official CDC webpage. Our model can consider the effect of many complexities of the epidemic process and yet be simplified to a few parameters that are learned using fast linear regressions. Therefore, our approach can learn and generate forecasts extremely quickly. On a 2 core desktop machine, our approach takes only 3.18s to tune hyper-parameters, learn parameters and generate 100 days of forecasts of reported cases and deaths for all the states in the US. The total execution time for 184 countries is 11.83s and for more than 3000 US counties is around 30s. For around 20,000 locations data for which are made available by Google, our approch takes around 10 mins. Despite being fast, the accuracy of our forecasts is on par with the state-of-the-art as demonstrated on the evaluation page. For country-level and US state-level projections, additional capabilities have been included in the model - currently, the model accounts for vaccinations and all variants as per outbreak.info.

Web Interface and Visualization

Our web-interface provides the following

1. Our US state-level and global country-level forecasts here
2. Our forecasts for around 20,000 location covering Admin 0-2
3. Weekly Highlights
4. Comparison against other public forecasts

Our papers

1. Full modeling details and comparisons: https://arxiv.org/abs/2007.05180
2. Identifying Unreported Cases; Accepted in KDD 2020): https://arxiv.org/abs/2006.02127
3. Initial Modeling: https://arxiv.org/abs/2004.11372

Presentations/Seminars

1. Lightning talk presenting the status (October): https://www.youtube.com/watch?v=ll6k8wlxOFo
2. Webinar describing our intial approach (April): https://www.youtube.com/watch?v=dBye3euqlKc

Acknowledgement

This work is supported by National Science Foundation Award No. 2027007 (2020-2021) and Award No. 2135784 (2021-2022).