covid-chicago-mobility

A UChicago class project aiming to predict reduction in mobility at the census block level as a result of the COVID-19 stay-at-home order in Chicago. Across Chicago there have been stark differences in the impact of COVID-19 in different neighborhoods (see Illinois Department of Public Health COVID-19 Statistics). Predicting reductions in mobility at the block-group level would allow for more specific neighborhood level policies to put in place, in order to effectively mitigate the impact of COVID-19.

Here we use a random forest model to predict reduction in movement at the neighborhood level in Chicago as a result of the first part of the stay-at-home order in Chicago (21st March - 1st May).

Data

This project uses cellphone mobility data from Safegraph combined with ACS census data at the census block group level in Chicago.

The mobility targets are created using Safegraph social distancing metrics data

The features for the model are in two parts:

1. ACS demographic data

2. SafeGraph Places Data: we created a count of each of the top 10 categories of places in the SafeGraph data (e.g. Grocery stores) still being visited in Chicago since the stay-at-home order for each census block group.

Running instructions

1. Set up

Set up a virtual environment (e.g. with virtualenv)and install package requirements from requirements.txt

2. Creating datasets

• All files for creating the target and features datasets are in the create_datasets directory
• All created files are saved within data directory
• All raw safegraph data (used by the modules in create_datasets is saved within data/raw

Step 1: filtering SafeGraph data to Cook County/Chicago, Illinois

create_datasets/safegraph_data_etl.py : for each of safegraph social_distancing, coreplaces and weekly_patterns data, takes raw data folders and filters all to Cook County/Chicago data into one filtered file

N.B. in order to run this, you first need access to the SafeGraph data, and to follow their instructions to download the raw data for each of the three datasets.

Step 2: create targets from SafeGraph social distancing data

run jupyter notebook: create_datasets/safegraph_targets/safegraph_targets.ipynb

This notebook reads in safegraph social distancing data, aggregates/computes targets, and plots histogram distributions of the targets.

Step 2: create features from ACS data

run create_datasets/census_features/generate_census_features.py

Step 3: create features from safegraph places data

run create_datasets/places_features/generate_places_features.py

3. Data exploration

data-analysis.ipynb - notebook to explore the features and targets. Creates choropleth maps of target and feature distribution across census block groups in Chicago. Map images are saved in './figure_outputs'

4. Running the models

COVID_model.ipynb - notebook runs a regularised regression model and a random forest model for each of our two targets.