Version 1.4.X

Version 1.4.1

Deployed: 13th June 2022



  • The BoxCoxTransformer has now inverse_transform functionality (Sangam)

  • Transformers now check for duplicated variable names entered in the init variables parameter (Soledad Galli)

Bug fixes


Version 1.4.0

Deployed: 9th June 2022


In this release, we fix a major bug that was preventing you guys from using the new module introduced in version 1.3: timeseries. We basically, forgot to add the __init__ file and for some reason, we did not pick up this error from our development environments. Thank you Brice for reporting this very important issue.

In addition, we updated our code base to work with the latest release of Scikit-learn (1.1.1) and pandas (1.4.2), which means that like Scikit-learn, we no longer support Python 3.7.

We are delaying the complete deprecation of MathematicalCombination, CombineWithFeatureReference, and CyclicalTransformer to our next release (1.5), as this release is a bit short notice, to give you more time to adapt your code bases.

In addition, we’ve added a new transformer, a number of new badges and made some enhancements to our code base.

I am very happy to announce that for this release, we had a number of contributions from first time contributors. Thank you so much for your support!

Thank you so much to all contributors to this release for making it possible!

New transformers

  • ArcsinTransformer: transforms variables with the arcsin transformation: arcsin(sqrt(x)) (tomtom-95)

Bug fixes

  • The SklearnTransformerWrapper now accepts categorical variables when used with the FunctionTransformer (Fernando Barbosa)

  • Added init file to allow import of time series module (Soledad Galli)



  • We are extending the complete deprecation of MathematicalCombination, CombineWithFeatureReference and CyclicalTransformer to version 1.5

Code improvements

  • Improved message returned when y is not what expected (following sklearn 1.1.1) (Soledad Galli)

  • Introduced check for some selectors to ensure user passes more than 1 variable (Soledad Galli)

For developers

  • We broke down base categorical classes into MixIns (Soledad Galli)

  • Accommodated lack of future pandas support for sets as indexers (Soledad Galli)