MLQA

A package to perform QA for Machine Learning Models.

Introduction

MLQA is a Python package that is created to help data scientists, analysts and developers to perform quality assurance (i.e. QA) on pandas dataframes and 1d arrays, especially for machine learning modeling data flows. It's designed to work with logging library to log and notify QA steps in a descriptive way.

Installation

You can install MLQA with pip.

pip install mlqa

MLQA depends on Pandas and Numpy and works in Python 3.5+.

Quickstart

You can easily initiate the object and fit a pd.DataFrame.

>>> from mlqa.identifiers import DiffChecker
>>> import pandas as pd
>>> dc = DiffChecker()
>>> dc.fit(pd.DataFrame({'mean_col':[1, 2]*50, 'na_col':[None]*50+[1]*50}))

Then, you can check on new data if it's okay for given criteria. Below, you can see data with increased NA count in column na_col. The default threshold is 0.5 which means it should be okay if NA rate is 50% more than the fitted data. NA rate is 50% in the fitted data so up to 75% (i.e. 50*(1+0.5)) should be okay. NA rate is 70% in the new data and, as expected, the QA passes.

>>> dc.check(pd.DataFrame({'mean_col':[1, 2]*50, 'na_col':[None]*70+[1]*30}))
True

See more examples at Documentation/Quickstart. You can also read the full documentation here.

Tests

Tests are written with unittest and can be located in the tests folder. There are also some tests in docstring to be run by doctest.

License

MIT