We provide a Python package called stability (the stability folder of this link ).
The package includes all tools of the paper:
"On the Stability of Feature Selection.". Nogueira, Sechidis and Brown. Journal of Machine Learning Reasearch (JMLR). 2017. (Accepted for publication with minor corrections, available online soon)including the functions to compute:
- Stability estimates.
- The variance of the estimates.
- Confidence intervals for the population stability.
- A hypothesis test allowing to compare the population stabilities of two feature selection procedures.
- A hypothesis test to check whether a population stability is greater than a given threshold.
- A demo in HTML (CLICK HERE) showing how to use this package and providing illustratating its use on real-data scenarios. You can also download the Jupyter Notebook.