About

This Typescript library provides a 1d persistent homology algorithm usefull for analyzing time series data. Implementation based on: Persistent homology in TDA

Note: calculations are computed using JavaScript primitive numbers, so floating point errors are to be expected.

Setup

1. Install NodeJS LTS
2. Install Yarn
3. Install dependencies

   yarn install
   

4. Build typescript code

   yarn build
   

Debugging

Enable debug logs

export DEBUG=ph:*

Running example/tests

After following the setup instructions, run teh following commands:

yarn test

This test runs a small example signal that can be traced by hand.

yarn csv-test ./test/owid-covid-data.csv 1

This script loads a csv file COVID-19 by country by day. It then computes the persistence of each total-cases data set, and the persistence of the first derivitive of each dataset. It then compares the distance between the persistence of each country and searches for the most similar country (first derivitive with smalles total distance).

The output.json file contais 2 main sections. Each Country is a Key in the map, which contains the preprocessed signal _processedData, its derivitive signal _processedDerivativeData, the persistence of both signals _ph and _phDerivative, and a map of the distance between each other country. The 2nd sction _comparison is the most similar country.

Sample data taken from: https://github.com/owid/covid-19-data/tree/master/public/data

Notes

Graphic created in matlab