dal3006 / ecfs

A Python implementation of the Ranking to Learn: Feature Ranking and Selection via Eigenvector Centrality paper.

Feature Ranking and Selection via Eigenvector Centrality

ECFS is a Python implementation of the Feature Ranking and Selection via Eigenvector Centrality algorithm by Giorgio Roffo and Simone Melz. Please check the code for full references.

Installation

Windows users can run the following:

git clone https://github.com/OhadVolk/ECFS.git
cd ECFS
python setup.py install

Linux users can run the following:

git clone https://github.com/OhadVolk/ECFS.git
cd ECFS
sudo python setup.py install

Usage

Check the Example.ipynb notebook for more details.

from ec_feature_selection import ECFS

# Create an instance, select top 10 features
ecfs = ECFS(n_features=10)
# Fit and Transform the training data
X_train_reduced = ecfs.fit_transform(X=X_train, y=y_train, alpha=0.5, positive_class=1, negative_class=0)
# Transform the test data
X_test_reduced = ecfs.transform(X_test)

License

MIT