Version 1.2.X

Version 1.2.0

Deployed: 4th January 2022

Contributors

In this big release, we add 3 new transformers, we expand the functionality of existing classes, we add information about citing Feature-engine and we expand the documentation with a new look, extended user guide with examples, and more details on how to contribute to the project.

Thank you so much to the contributors for making this massive release possible!

Thank you to reviewers Nicolas Galli and Chris Samiullah for useful advice on various PRs.

New transformers

  • DatetimeFeatures: extracts date and time features from datetime variables (Edoardo Argiolas)

  • DropHishPSIFeatures: finds and drops features with high population stability index (gverbock)

  • Matchvariables: ensures that the same variables observed in the train set are present in the test set (Thibault Blanc)

Enhancements

  • The Winsorizer can now add binary variable indicators to flag outlier values (David Cortes)

  • The DropMissingData now allows to drop rows based on % of missing data (Kevin Kurek)

  • Categorical encoders can now either raise a warning or an error when encoding categories not seen in the train set (Morgan Sell)

  • The ArbitraryDiscretiser can now either raise a warning or an error when values fall outside the limits entered by the user (Morgan Sell)

  • CombineWithReferenceFeature and MathematicalCombination have now the option to drop the original input variables after the feature creation (Edoardo Argiolas)

Bug fixes

  • All Encoders are now able to exclude datetime variables cast as object or categorical when searching for categorical variables automatically (Edoardo Argiolas)

  • All transformers will now raise an error when users pass an empty list to the variables parameter (Edoardo Argiolas)

  • All transformers now check the variable type when user passes a single variable to the variables parameter (Edoardo Argiolas)

Documentation

  • We changed the template to pydata (Soledad Galli)

  • We split the information about transformers into a user guide and an API (Soledad Galli)

  • The API documentation shows how to use the transformers (Soledad Galli)

  • The user guide expands the API docs with plenty of examples and tips on when and how to use the transformers (Soledad Galli)

  • We expanded the contribute section with plenty of details on how to make a contribution and how to check your code is top notch (Soledad Galli)

  • You can now sponsor Feature-engine (Soledad Galli)

  • You can now cite our JOSS article when using Feature-engine (Soledad Galli)

  • We added plenty of examples on how to use the new class DropHighPSIFeatures (gverbock)

  • We included various examples on how to extract date and time features using the new DatetimeFeatures class (Edoardo Argiolas)

  • We included examples on how to use the new class MatchVariables (Thibault Blanc)

  • We added a Jupyter notebook with a demo of the new DatetimeFeatures class (Edoardo Argiolas)

  • We added a Jupyter notebook with a demo of the new DropHighPSIFeatures class (Soledad Galli)