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] and Encoding high-cardinality string categorical variables [2].

dirty_cat requires:

• Python (>= 3.5)
• NumPy (>= 1.8.2)
• SciPy (>= 1.0.1)
• scikit-learn (>= 0.20.0)

Optional dependency:

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

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

• Spark ML: https://github.com/rakutentech/spark-dirty-cat