About This Project

This project was an accompanying open source module that I made to generate the simulations that were on my Covid-19 tracking website.

The simulation results are available here. Note the website takes about 30 seconds to load cause its currently on a free server.

At first I was interested in seeing how I can test the effectiveness of various measures against mitigating disease spread. I watched Steve Brunton's series on simple mathematical modelling of disease spread and I started writing some code to get some results. I wrote a module that simulates a bunch of nodes that move randomly inside a confined space. I assigned values to speed of movement , probability of infection, radius of infection, etc. that I tried my best to estimate from doing some research. The program then simulates disease spread until the infected number of nodes reach zero where there are 4 possible states to a node at a given time: infected, diseased, recovered, susceptible. Obviously, this is a simple model of disease spread, but I intended for this module to be something small and easy to use to produce some meaningful simulations.

Technical Information

One obvious issue I ran into while creating the module is the performance of the animation. Part of determining the disease spread is determining the status of each node at each time stamp. The most perfomance consuming task was checking for each node, which are the neighboring nodes that are within an infectious radius. To implement that this checking method, I used a K-dimensional tree to store the node positions and then queried for the closest nodes using that.

I used matplotlib to generate all my animations. One thing I tried is saving and exporting the pandemic data so I can use a front-end animation library such as D3.js to create the animations on the client-side. The idea was to run the pandemic module on a server were users can create there own simulations and the data would be returned through a REST API where a client-side animation library does the work. I ran into some performance issues attempting to run each users' requested animation on my 7$ Heroku hobby server. I also wanted to couple the animations with the actual simulation module so I just genereated all the animations in Python and exported them as video files.

Status of Project

I have not made edits to the project recently. Some things I might do down the road is improve the performance of simulation to allow for bigger sized simulations. I can also add more realistic factors that impact disease spread.

Another thing I wanted to do is maybe do an accompanying Javascript library that performs simulations on client side which would make the usage more usage/developer friendly.