[AAAI-21] Steering a Historical Disease Forecasting Model Under a Pandemic: Case of Flu and COVID-19

Publication

Implementation of the paper "Steering a Historical Disease Forecasting Model Under a Pandemic: Case of Flu and COVID-19" published in AAAI-21.

Authors: Alexander Rodríguez, Nikhil Muralidhar, Bijaya Adhikari, Anika Tabassum, Naren Ramakrishnan, B. Aditya Prakash

Venue: AAAI Conference on Artificial Intelligence (AAAI-21)

Pre-print: https://arxiv.org/abs/2009.11407

Appendix: LINK

Requirements

Use the package manager conda to install required Python dependencies. Note: We used Python 3.7.

conda env create -f requirements.yml

Training

The following command will train and predict for all regions from epidemic week 9 to 15:

python ./main.py --start_week 9 --end_week 15

You can set up your own model hyperparameter values (e.g. learning rate, loss weights) in the file ./experiment_setup/feature_module/model_specifications/global_recurrent_feature_model.json.

Evaluation

To evaluate the results, go to evaluate.py and change line 71 for the name of results file (saved in folder rmse_results). Then, run.

python ./evaluate.py

Contact:

If you have any questions about the code, please contact Alexander Rodriguez at arodriguezc[at]gatech[dot]edu and/or B. Aditya Prakash badityap[at]cc[dot]gatech[dot]edu