RuleCOSI

Rule extraction COmbination and SImplification from classification tree ensembles

RuleCOSI is a machine learning package that combine and simplifies tree ensembles and generates a single rule-based classifier that is smaller and simpler. It was developed in the Industrial Artificial Intelligence Laboratory (IAI) at Kyung Hee University by (Josue Obregon). The implementation is compatible with scikit-learn.

Note

The github repository of this project is going to be open to the public soon (April 2021) with the alpha version of the library.

Indices and tables

About

If you use rulecosi in a scientific publication, we would appreciate citations to the following paper:

@article{obregon2019rulecosi,
  title={RuleCOSI: Combination and simplification of production rules from boosted decision trees for imbalanced classification},
  author={Obregon, Josue and Kim, Aekyung and Jung, Jae-Yoon},
  journal={Expert Systems with Applications},
  volume={126},
  pages={64--82},
  year={2019},
  publisher={Elsevier}
}

The algorithm works with different type of ensembles and it uses the implementations provided by the sklearn package. The supported tree ensemble types are:

  1. BaggingClassifier

  2. RandomForestClassifier

  3. GradientBoostingClassifier

  4. XGBClassifier

  5. LGBMClassifier

  6. CatBoostClassifier

For more information you can check the usage in the docstrings or the examples folder of this repository.

References:

1

Obregon, J., Kim, A., & Jung, J. Y. (2019). RuleCOSI: Combination and simplification of production rules from boosted decision trees for imbalanced classification. Expert Systems with Applications, 126, 64-82.