covid19-on-graphs

This repo contains datasets for spatio-temporal prediction of COVID-19 distribution over the US. The US is represented as a graph where the nodes are associated with states, and edges represent the adjacency of the states.

How to use it?

1. Download raw data from the NY times using this link and put it to ./data.
2. Install dependencies:

pip install networkx==2.8.8 pandas==1.3.5

3. Use datasets.covid_adjacencies_graph_dataset() to get graph and temporal data.

import datasets
graph, temp_data = datasets.covid_adjacencies_graph_dataset()

4. Enjoy your experiments!

Format

Graph

A graph is stored as nx.Graph object. Read more about it in networkx documentation.

Temporal data

The data are aggregated weekly and in the following format:

temp_data[date][state] = {
    "deaths": INT, # The number of deaths during the week
    "cases": INT,  # The number of cases during the week
    "deaths_normalized": FLOAT,  # The number of deaths normalized by the state population
    "cases_normalized": FLOAT,   # The number of cases normalized by the state population
}