This repository contains Monte Carlo simulation tools for corona virus spread predictions
Setup:
pip install -r requirements.txt
Run:
python main.py
Computational resources used (simulation with a single city with 6 million inhabitants in 10 different populations):
- RAM: <500 MB
- CPU: 1.6 s per simulated day
Models revolve around populations which are represented by the PopulationBase class (tools/population.py), which:
- keeps track of health states of its members
- keeps track of when its members contracted the virus
- immunity of members towards the virus
- which members are still alive
- random selection of subsamples
- which members have recovered from the illness
- define interaction patterns for individuals (interactions withing households, random interactions within a community, etc.)
See examples/basic.py for example usage
Abstract geographically defined concentrations of human population (e.g. cities) and simulate daily interactions among different populations. Can hold multiple populations with different parameters and interaction patterns (e.g. children, working class, elderly people etc.). A transmission matrix can be defined to simulate different transmission probabilities between different populations.
Defined in tools/population_centre.py
The class stores results for each timestep.
See examples/single_city.py for example usage
Viruses are represented by subclasses of the Virus class, which define:
- mean period between infection and recovery
- standard deviation of the period between infection and recovery
- transmission probability of the given virus (R), which can depend on multiple factors
- mortality of the given virus
n.b. only a single virus can spread among populations since they do not keep track of the type of virus
Need to tune the model parameters (R, random interaction distributions, transmission probabilities, ...). This can be done by tuning the parameters so that a medium sized single city model would match results of https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf and heuristically define social interaction patterns for typical Slovak populations.
- add configuration files for tunable parameters
- add data readers for various sources of municipal data (town populations, migration matrixes, ...)
- prepare a model with all Slovak cities, towns and villages
- prepare plotting / result visualization for geographical simulations
- run on multiple CPU cores