This repository contains the code used for the paper "Bayesian calibration of stochastic agent based model via random forest". It contains hospitalization and death data produced by the CityCOVID agent based model. It also provides code to train a Random Forest surrogate model for CityCOVID hospitalizations and deaths, calculate a Bayesian estimate of parameters from CityCOVID using this surrogate, and then produce plots and data of this calibration.
All dependencies can be installed by running make install_deps.
Specifics can be found below.
The code in this repository makes use of the following python packages:
matplotlibnumpypandasproperscoringpymcmcstat(from updated version here)seabornscikit-learn
These can easily be installed by running make install_python if you have conda or pip already on your system.
Alternatively, they can be installed with pip via pip install -r requirements.txt or with conda/mamba via conda env create -f environment.yml.
The code in this repository makes use of the following R packages:
ggdistcodamcgibbsitscoringutils
These can easily be installed by running make install_r as long as R is already installed on your system.
Reproduction of the results and plots can most easily be done by running make all_surrogate followed by make all_calibration and make compare.
The above runs various scripts from scripts/ which make use of some utility functions in src/ and data from data/ to output files into results/ and plots into plots/.
See make help or make for more information on individual scripts.