Skip to content

Latest commit

 

History

History
128 lines (94 loc) · 4.47 KB

README.md

File metadata and controls

128 lines (94 loc) · 4.47 KB
Raw

pyepidemics

logo

Open-source epidemics modeling Python library

pyepidemics is a Python library to easily manipulate epidemiological models (SIR, SEIR, COVID19), forecast, and test policy scenarios. Main features are:

  • Creation of classical compartmental models (SIR, SEIR, SEIDR, etc...)
  • Creation of COVID19 related model (with ICU and different levels of symptoms)
  • Creation of custom compartmental model
  • Easy extend to compartmental model on different levels (by age, by region, by age and region, etc...)
  • Implementation of contact matrices
  • Implementation of policies (lockdown, tracing, testing, etc...)
  • Calibration of epidemiological parameters on real-world data using Bayesian optimization
  • Data helpers to get up-to-date data (cases, beds availability, population contact matrices) (NB as of today only for France cases)
  • Simple curve fitting algorithms
  • Economics modeling on consumption and production

This library is under active development, please contact Théo Alves Da Costa if you need more information and would like to contribute

Contributors

The library has been initially developed by Ekimetrics under the coalition of French AI companies CoData to help French government response to the COVID19 pandemics. Main contributors are:

  • Théo Alves Da Costa, Ekimetrics
  • Emilie Rannou, Ekimetrics
  • Léo Grosjean, Ekimetrics
  • Pierre Carles, Ekimetrics
  • Nicolas Chesneau, Ekimetrics
  • Marianne Chehade, Ekimetrics
  • Jean-Baptiste Remy

Installation

Install using pip

We recommend to create a virtual environment first, then you can install the library using the command

pip install pyepidemics

The repo is on PyPI at https://pypi.org/project/pyepidemics/

Install from source

You can also install the repository by cloning it locally and using it either locally or installing the wheel by calling first

python setup.py sdist bdist_wheel

Documentation

Documentation is available at https://ekimetrics.github.io/pyepidemics

Folder structure

- pyepidemics/          --------- Python library centralizing source code
    - dataset/          --------- Helpers to get up-to-date COVID19 datasets
    - models/           --------- Compartmental models code
    - params/           --------- Epidemiological parameters calibration optimizers
    - policies/         --------- Scenarios implementation
    - visualization/    --------- Visualization helpers (chloropleth maps using pydeck)
- data/                 --------- Local datasets if needed
- docs/                 --------- Documentation and tutorial notebooks
- notebook/             --------- Development notebooks
- references/           --------- Reports and research papers
- scripts/              --------- Automation scripts on calibration
- requirements.txt      --------- Python requirements

Quickstart

Creating a SIR model

# Import library
from pyepidemics.models import SIR

# Let's take approximate parameters during COVID19 epidemics
N = 67e6
beta = 3.3/4
gamma = 1/4

# Instantiate model with epidemiological parameters
sir = SIR(N,beta,gamma)

# Solve for one infected case for 100 days starting from 2020-01-24
states = sir.solve(1,n_days = 100,start_date = "2020-01-24")

# Visualize epidemic curves using matplotlib (plotly available)
states.show(plotly = False)

Creating a custom compartmental model

Here is the example to create a model like the SIR described above

from pyepidemics.models import CompartmentalModel


class SIR(CompartmentalModel):
    def __init__(self,N,beta,gamma):
        
        # Define compartments name and number
        compartments = ["S","I","R"]
        super().__init__(compartments)

        # Parameters
        self.N = N # Total population
        self.beta = beta # How many person each person infects per day
        self.gamma = gamma # Rate of infection, duration = 1/gamma
        
        # Add transition
        self.add_transition("S","I",lambda y,t: self.beta * y["S"] * y["I"] / self.N)
        self.add_transition("I","R",lambda y,t: self.gamma * y["I"])

Contribution guidelines

WIP

References

License

MIT License