Unsupervised Features Ranking via Coalitional Game Theory for Categorical Data

Welcome to the code for our paper, Unsupervised Features Ranking via Coalitional Game Theory for Categorical Data, published at DaWaK 2022. We encourage you to read the full paper.

Citation

If you found this work useful, please cite our paper:

@inproceedings{balestraUSV,
  author    = {Chiara Balestra  and
               Florian Huber and
               Andreas Mayr and
               Emmanuel M\"uller},
  title     = {Unsupervised Features Ranking via Coalitional Game Theory for Categorical Data},
  booktitle = {DaWaK 2022},
  publisher = {Springer},
  year      = {2022}
 }

Example and code

The shapley_calculation.py contained the Shapley values implementation and feature_selection.py includes the SVFR and SVFS algorithms. We provide an example of the use of the code in _example.py referring to a synthetic data set.

In order to run the code use

python example.py --_algorithm='SVFR' --_type='full' --_epsilon=0.6 --_approx=3 --_subsets_bound=2

and change the parameters as desired.

Requirements

Code tested under:

• python 3.7.6
• numpy 1.18.5
• pandas 1.4.0

External librarys used:

• pyitlib 0.2.2 (pip install pyitlib)

Questions

You can reach out to chiara.balestra1@gmail.com with any question