FImdlp

Discretization algorithm based on the paper by Usama M. Fayyad and Keki B. Irani

Multi-Interval Discretization of Continuous-Valued Attributes for Classification Learning. In Proceedings of the 13th International Joint Conference on Artificial Intelligence (IJCAI-95), pages 1022-1027, Montreal, Canada, August 1995.

Installation

From PyPI

pip install FImdlp

From source (development)

The project no longer relies on a git submodule — the C++ sources live under src/cpp/. A regular clone is enough:

git clone https://github.com/doctorado-ml/FImdlp.git
cd FImdlp
make deps      # install the [dev] extras (build, twine, pip-audit, black, flake8, coverage)
make install   # editable install; compiles the Cython/C++ extension in place

Run make help to list every available target.

Quick start

from sklearn.datasets import load_iris
from fimdlp.mdlp import FImdlp

X, y = load_iris(return_X_y=True)
clf = FImdlp().fit(X, y)

# Discretize
X_disc = clf.transform(X)

# Inspect cut points: [vmin, c1, ..., cn, vmax] per feature
for f, cuts in enumerate(clf.get_cut_points()):
    print(f"feature {f}: {cuts}")

Constructor parameters:

Make targets

Running the samples

Python sample

make sample_py
# equivalent to:
#   cd samples && python sample.py iris

Other options:

python samples/sample.py iris            # default settings
python samples/sample.py iris -c 2       # cap intermediate cut points to 2
python samples/sample.py iris -m 3       # cap recursion depth to 3
python samples/sample.py iris -n 25      # set min_length to 25
python samples/sample.py -h              # full option list

C++ sample

make sample_cpp
# equivalent to:
#   cd samples && cmake -B build -S . && cmake --build build && cd build && ./sample -f iris

Other options:

cd samples/build
./sample -f iris -c 2     # cap intermediate cut points to 2
./sample -f glass -m 3    # change dataset and depth
./sample -h               # full option list

Based on

https://github.com/rmontanana/mdlp