This repository provides the official implementation of ADABORD, a novel AdaBoost-based ensemble method designed for ordinal classification tasks. This repository contains the code accompanying the scientific article:

AdaBORD: Adaptive Boosting for Ordinal Classification with Decision Trees

• 🚀 Built on top of scikit-learn, extending it with ordinal-aware boosting strategies.
• 🚀 Introduces a novel AdaBoost formulation tailored to exploit the natural order of class labels.
• 🚀 Designed for reproducibility — all experiments from the article can be replicated with the provided code.

✨ ADABORD is a top-performing method on the TOCUCO benchmark, the standard repository for ordinal classification comparison.

ADABORD relies on a modified version of scikit-learn that introduces ordinal splitting criteria. Install it directly from the source branch:

pip install git+ssh://git@github.com/RafaAyGar/scikit-learn.git@feature_ogini_split

Or with uv:

uv pip install "scikit-learn @ git+ssh://git@github.com/RafaAyGar/scikit-learn.git@feature_ogini_split"

If you use ADABORD in your research, we would appreciate a citation for the following work:

citation not available yet