Source code for paper "Reinforcement Learning Enhances the Experts: Large Scale COVID-19 Vaccines Allocation with Multi-factor Contact Network" (KDD'22)

• These files are uploaded to Github and supplementary materials in the Microsoft CMT at the same time.
• Due to file size limitations, the folder '/data' only serves as a PLACEHOLDER and the data needed to be downloaded and put into it follow the instruction below.

Implement steps

Here are the implementation steps for Atlanta experiments as an example.

1. Download SafeGraph Open Census Data from https://docs.safegraph.com/docs/open-census-data and replace the folder '/data/safegraph_open_census_data'. Actually, we only need one file in this dataset: '/data/safegraph_open_census_data/data/cbg_b01.csv'.
2. Download the IDs of the CBGs in Atlanta 'Atlanta_Sandy_Springs_Roswell_GA_cbg_ids.csv' from https://covid-mobility.stanford.edu//datasets/ [1] and place it as '/data/Atlanta/Atlanta_Sandy_Springs_Roswell_GA_cbg_ids.csv'.
3. Download the raw contact network data in Atlanta 'Atlanta_Sandy_Springs_Roswell_GA_2020-03-01_to_2020-05-02.pkl' from https://covid-mobility.stanford.edu//datasets/ [1] and place it as '/data/Atlanta/Atlanta_Sandy_Springs_Roswell_GA_2020-03-01_to_2020-05-02.pkl'.
4. Get the average dwelling time and floor space of POIs in Atlanta '/data/Atlanta/parameters_Atlanta.csv' with the method in [1].
5. Get the CBG specific death rate in Atlanta '/data/Atlanta/cbg_death_rates_original_Atlanta' with the method in [2].
6. Run '/code/data_process.py' and get the processed contact network data in Atlanta '/data/Atlanta/Atlanta_Sandy_Springs_Roswell_GA_2020-03-01_to_2020-05-02_processed.pkl'.
7. Run '/code/aggregate.py' and get the community division indexes in Atlanta '/data/Atlanta/community_index_fluid.npy'.
8. Run '/code/process_aggregate.py' and get the aggregated contact network data in Atlanta '/data/Atlanta/Atlanta_Sandy_Springs_Roswell_GA_2020-03-01_to_2020-05-02_processed_aggregate.pkl' and the corresponding sizes and age structures of the communities '/data/Atlanta/aggregate_sizes.npy', '/data/Atlanta/aggregate_ages.npy'
9. Run '/code/train.py' for model training and the model are saved in '/model/Atlanta' for testing and analyzing.
10. Run '/code/baseline.py' for baseline results roll out.

Other Python files in '/code' not mentioned above are imported by the mentioned ones.

Reference

[1] Serina Y Chang*, Emma Pierson*, Pang Wei Koh*, Jaline Gerardin, Beth Redbird, David Grusky, and Jure Leskovec. "Mobility network models of COVID-19 explain inequities and inform reopening". Nature, 2020.

[2] Lin Chen, Fengli Xu, Zhenyu Han, Kun Tang, Pan Hui, James Evans, and Yong Li. "Strategic COVID-19 vaccine distribution can simultaneously elevate social utility and equity". arXiv preprint, 2021.