Simple SEIR model Python script for the COVID-19 pandemic with real world data.
Purpose is to easily be able to experiment and better understand what is happening currently and what might happen in the near future.
What I learned: Activating containment measures early can save lives.
There is another branch in this repository that is about blood type (ABH, ABO, AB0) extended SEIR modeling and correlating model data vs real world data: https://github.com/coronafighter/coronaSEIR/tree/ABH
- Nice webgui: http://gabgoh.github.io/COVID/index.html
- Another great web based model: https://neherlab.org/covid19/
- rewrite of this project with sophisticated code: https://github.com/cfculhane/coronaSEIR
This is not a scientific or medical tool. Use at your own risk. BETA! There might be serious bugs.
- SEIR epidemic model
- Reduced R0 after a certain amount of days to account for containment measures.
- Delays to allow for lagging official data etc.
- hopefully easily readable code
- Real world data automatically updated every three hours from Johns Hopkins CSSE (https://github.com/CSSEGISandData/2019-nCoV) via https://github.com/ExpDev07/coronavirus-tracker-api
- country population data (https://github.com/samayo/country-json)
- can exclude countries, e.g. world without China
- calculate and chart R0
- ABH-Branch: experiments about blood type and COVID-19. Correlation of country models vs real data
- check out screenshots below
Needs Python 3.x installed. Tested on Ubuntu. Should run on Window and Mac, too.
# everything after the hash symbol # is a comment
$ git clone https://github.com/coronafighter/coronaSEIR # create local copy
$ pip3 install --upgrade numpy scipy matplotlib python-dateutil # might need sudo -H pip3 ...
$ python3 ./main_coronaSEIR.py
read data: 843103 bytes
r0: 5.20 r1: 1.30
doubling0 every ~3.4 days
day 60
Infected: 119363 0.2
Infected found: 1085 0.0
Hospital: 787 0.0
Recovered: 44584 0.1
Deaths: 17 0.0
data points for Italy: 40
first data: 2020-01-31
latest data: 2020-03-10 (you can update the data manually by running fetch_data.py)
$python3 ./world_data.py # to just plot current numbers
$python3 ./deaths_per_capita.py # might be a better measure than cases per capita
$ git pull origin master # update to latest version, might overwrite local changes
No GUI, you need to alter the script and run again to experiment.
Note: Make sure you got correct number for population and available ICU units for your country.
- change main_coronaSEIR to daily changes instead of cumulated numbers
- idea: compare optimal fit doubling time for all countries (and against countries blood type distribution)
- idea: calculate for all countries to compare how badly hit (days0 needs to be set manually)
- add data about lockdown starting times, also make it possible to enter as a date
- make R0 and days0 lists to be able to have more than two phases
- add data about intensive care units
- make parsing code cleaner / simpler
- be more precise in differentiation between hospitalization and ICU
- n ventilator patients separately?
- X automatic date offset
- X maybe find a better fit with lower R0 - the infectious time seems to be quite short?
- X add proper list of populations
- X use real dates instead of days?
Please note that I would prefer to stick to the code style of camelCase variables and snake_case function/method names to better distinguish the two. I am aware of PEP8.
I would like to keep the monolithic main file with interactive parameters, model function and output all in one. Code is kept rather simple on purpose to allow for good readability even for beginners (I know, some parts are still very ugly).
Let me know if you would like me to add your project or fork to this readme.
Based on:
https://github.com/ckaus/EpiPy
https://scipython.com/book/chapter-8-scipy/additional-examples/the-sir-epidemic-model/
API/Data: https://github.com/ExpDev07/coronavirus-tracker-api https://github.com/samayo/country-json
Formulas:
https://hal.archives-ouvertes.fr/hal-00657584/document
https://institutefordiseasemodeling.github.io/Documentation/general/model-seir.html
Parameters:
https://www.reddit.com/r/COVID19/comments/fbdzc1/coronavirus_epidemiology_metaanalysis_and/
https://www.reddit.com/r/COVID19/comments/fbxk43/update_open_source_simple_coronavirus_modeling/
timeline https://www.reddit.com/r/COVID19/comments/fd6lmg/infectionfatalityratio_ifr_of_covid19_is/
https://www.reddit.com/r/COVID19/comments/ffzqzl/estimating_the_asymptomatic_proportion_of_2019/
https://www.reddit.com/r/Coronavirus/comments/f8k2nj/why_sarscov2_is_not_just_the_flu_with_sources/
gamma: Generation time (serial interval): https://www.medrxiv.org/content/10.1101/2020.03.05.20031815v1
also: http://www.cidrap.umn.edu/news-perspective/2020/03/short-time-between-serial-covid-19-cases-may-hinder-containment
Relationship hospitalized/ICU/death https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf
r0 ~ 3.0 https://www.newscientist.com/article/2238578-uk-has-enough-intensive-care-units-for-coronavirus-expert-predicts/
r1 (thanks paus-5!) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3539694
sigma: Infection occurs before symptoms (Drosten): https://www.medrxiv.org/content/10.1101/2020.03.08.20032946v1.full.pdf
Number of asymptomatics: https://www.zmescience.com/medicine/iceland-testing-covid-19-0523/
Model with AB0 blood type classes: https://www.researchgate.net/figure/An-example-of-a-two-class-age-structured-SIR-model-with-parameters-for-a-typical-human_fig3_259825206
some more in the source code
MIT license