dirty-cat 0.0.2

Machine learning with dirty categories.

dirty_cat is a Python module for machine-learning on dirty categorical variables.

Website: https://dirty-cat.github.io/

For a detailed description of the problem of encoding dirty categorical data, see Similarity encoding for learning with dirty categorical variables [1].

Installation

Dependencies

dirty_cat requires:

• Python (>= 3.5)

• NumPy (>= 1.8.2)

• SciPy (>= 1.0.1)

• scikit-learn (>= 0.19.0)

Optional dependency:

• python-Levenshtein for faster edit distances (not used for the n-gram distance)

User installation

If you already have a working installation of NumPy and SciPy, the easiest way to install dirty_cat is using pip

pip install -U --user dirty_cat

References

[1]

Patricio Cerda, Gaël Varoquaux, Balázs Kégl. Similarity encoding for learning with dirty categorical variables. 2018. Accepted for publication in: Machine Learning journal, Springer.