dirty_cat
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. |