CorrectMatch

A thin Python wrapper around the Julia module CorrectMatch.jl, to estimate uniqueness from small population samples.

Installation

Install first Julia and CorrectMatch.jl, then this Python wrapper: pip install correctmatch

We use PyJulia to seemingly run Julia code from Python. Your Julia installation should be automatically detected, otherwise follow the instruction on the PyJulia documentation.

Usage

This module estimates the uniqueness of a population, on which multiple discrete attributes can be collected. For instance, the following array is a sample of 1000 rows for five discrete attributes:

>>> import numpy as np
>>> arr = np.random.randint(1, 5, size=(1000, 5))
>>> arr[:3, :]
array([[1, 1, 1, 3, 1],
       [3, 3, 2, 3, 3],
       [3, 3, 4, 3, 2]])

We can estimate the uniqueness of a population of 1000 individuals, or 10000 individuals, from this sample:

>>> import correctmatch
>>> correctmatch.precompile()  # precompile the Julia module

>>> correctmatch.uniqueness(arr)  # true uniqueness for 1,000 records
0.38

by fitting a copula model to the observed records:

>>> fitted_model = correctmatch.fit_model(arr)
>>> fitted_arr = correctmatch.sample_model(fitted_model, 1000)
>>> fitted_arr[:3, :]
array([[4, 2, 1, 4, 1],
       [4, 2, 3, 2, 3],
       [1, 3, 1, 3, 1]])
>>> correctmatch.uniqueness(fitted_arr)
0.393

In the demo/ folder, we have compiled more examples with real-world datasets.

License

GNU General Public License v3.0

See LICENSE to see the full text.

Patent-pending code. Additional support and details are available for commercial uses.