With the widespread use of machine learning to support decision-making, it is increasingly important to verify and understand the reasons why a particular output is produced. Although post-training feature importance approaches assist this interpretation, there is an overall lack of consensus regarding how feature importance should be quantified, making explanations of model predictions unreliable. In addition, many of these explanations depend on the specific machine learning approach employed and on the subset of data used when calculating feature importance.
A possible solution to improve the reliability and understandability of explanations is to combine results from multiple feature importance quantifiers from different machine learning approaches coupled with data re-sampling.
EFI toolkit implements this solution using:
- State-of-the-art information fusion techniques
- Fuzzy sets
The toolkit provides complete automation of the entire feature importance computation cycle.
- automatic training and optimisation of machine learning algorithms.
- automatic computation of a set of feature importance coefficients from ensemble of optimised machine learning algorithms and feature importance calculation techniques.
- automatic aggregation of importance coefficients using multiple decision fusion techniques.
- automatic generation of fuzzy membership functions that show the importance of each feature to the prediction task in terms of
low',moderate' and `high' importance as well as their levels of uncertainty.
- pip install dataframe-image
- pip install scikit-learn
- pip install tensorflow
- pip install silence-tensorflow
If you use EFI in scientific publications, we would appreciate citations.
EFI: A Toolbox for Feature Importance Fusion and Interpretation in Python Aayush Kumar, Jimiama Mafeni Mase, Divish Rengasamy, Benjamin Rothwell, Mercedes Torres Torres, David A. Winkler, Grazziela P. Figueredo Conference on Machine Learning, Optimization, and Data Science (2022)
Link to publication.
@article{kumar2022efi,
title={EFI: A Toolbox for Feature Importance Fusion and Interpretation in Python},
author={Kumar, Aayush and Mase, Jimiama Mafeni and Rengasamy, Divish and Rothwell, Benjamin and Torres, Mercedes Torres and Winkler, David A and Figueredo, Grazziela P},
journal={arXiv preprint arXiv:2208.04343},
year={2022}
}
Feature importance in machine learning models: A fuzzy information fusion approach Divish Rengasamy,Jimiama M.Mase, Aayush Kumar, Benjamin Rothwell, Mercedes Torres Torres, Morgan R.Alexander, David A.Winkler, Grazziela P.Figueredo* Neurocomputing Journal, 2022
Link to publication.
@article{rengasamy2022feature,
title={Feature importance in machine learning models: A fuzzy information fusion approach},
author={Rengasamy, Divish and Mase, Jimiama M and Kumar, Aayush and Rothwell, Benjamin and Torres, Mercedes Torres and Alexander, Morgan R and Winkler, David A and Figueredo, Grazziela P},
journal={Neurocomputing},
year={2022},
publisher={Elsevier}
}
**The toolkit and its description will be updated as new explainability techniques and machine learning models are implemented.**