🌳 Welcome to decision-trees-from-scratch

This framework provides a from scratch sklearn-based implementation of the CART algorithm for classification. Our implementation introduces notable differences compared to the existing sklearn DecisionTreeClassifier:

• 🚀 It is fully developed in python. This enables researchers to easily tweak the algorithm and experiment with it for research purposes.

• 🚀 Includes current state-of-the-art splitting criteria for ordinal regression tasks, such as Ordinal Gini or Weighted Entropy.

✨ Despite being developed from scratch, our implementation achieves same accuracy results as scikit-learn implementation.

⚙️ Installation

decision-trees-from-scratch requires a Python version of 3.9 or greater. Follow the next step for installation:

1. Download or clone the repository
2. On the root folder, run: pip install . We recommend to use the --editable argument to be able to modify the package on spot.

📚 Citation

If you enjoyed this framework, we would appretiate a citation for the following work:

@misc{ayllóngavilán2024splittingcriteriaordinaldecision,
      title={Splitting criteria for ordinal decision trees: an experimental study}, 
      author={Rafael Ayllón-Gavilán and Francisco José Martínez-Estudillo and David Guijo-Rubio and César Hervás-Martínez and Pedro Antonio Gutiérrez},
      year={2024},
      eprint={2412.13697},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2412.13697}, 
}