BFMDT

Bi-directional Fusing Monotonic Decision Trees.

Python implementation of the algorithm from:

"Fusing Monotonic Decision Tree Based on Related Family" IEEE Transactions on Knowledge and Data Engineering, 2024

Setup

# Clone
git clone git@github.com:nhinbm/BFMDT.git
cd BFMDT

# Create virtual environment
python -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements/base.txt

# For development
pip install -r requirements/dev.txt

# For running experiments
pip install -r requirements/experiment.txt

Usage

from bfmdt import BFMDTClassifier

clf = BFMDTClassifier(sigma="auto", delta=0.01, max_reducts=50)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

Run Experiments

# Run on a single dataset
python experiments/run_benchmarks.py --dataset wine

# Run on all 18 datasets
python experiments/run_benchmarks.py --dataset all

# Custom parameters
python experiments/run_benchmarks.py --dataset wine --n_folds 5 --output_dir results/

Run Tests

pytest tests/

Docs

• PIPELINE.md -- Algorithm pipeline
• INTERFACES.md -- Module interfaces
• STRUCTURE.md -- Project structure and datasets