python-feature-test

Python implementation for the Discriminant Feature Test (DFT) and Relevant Feature Test (RFT) in sklearn style.

Requirements

numpy==1.21.5
scikit-learn==1.0.1

Binary Classification

Adopt the binary cross-entropy (BCE) loss for feature selection. Implementation of the loss function follows the equation here.

dft = FeatureTest(loss='bce')

Regression

Adopt the root mean-squared-error (RMSE) as the loss function for feature selection. Consider turning on outliers in fit() if there are some anomalies in the datasets. Implementation of the loss function follows the equation here.

rft = FeatureTest(loss='rmse')

Multi-class Classification

Use the cross-entropy (CE) loss for multi-class classification (n_class > 2) problems.

dft = FeatureTest(loss='ce')

Multi-label Classification

Multi-label classification can be formulated as n_class independent binary classification. Consider using BCE loss to select features for the multi-label classification problems.

Demo

The basic usage of the package is included in the demo. More demos will be available soon.

Citation

Please consider citing the original feature test paper if you find this code useful.

@article{yang2022supervised,
  title={On supervised feature selection from high dimensional feature spaces},
  author={Yang, Yijing and Wang, Wei and Fu, Hongyu and Kuo, C-C Jay and others},
  journal={APSIPA Transactions on Signal and Information Processing},
  volume={11},
  number={1},
  year={2022},
  publisher={Now Publishers, Inc.}
}