Mean Normalisation Scaling

[VascoSch92]

First version of the MeanNormalizationScaling as discussed in #763

I create a new module scaling as discussed.

Probably, you will also add new scaling in this module.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.00%. Comparing base (5dfceb8) to head (b001017).
Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #806      +/-   ##
==========================================
+ Coverage   97.98%   98.00%   +0.01%     
==========================================
  Files         107      109       +2     
  Lines        4320     4350      +30     
  Branches      857      709     -148     
==========================================
+ Hits         4233     4263      +30     
  Misses         54       54              
  Partials       33       33              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[solegalli]

Hey @VascoSch92

This is looking really good. Thank you for the first draft.

I think we could tidy it a bit so that we don't loop neither in fit nor in transform.

Could you take a look?

Thank you!

[VascoSch92]

Hey @solegalli

I addressed your comments.

Please let me know what you think :-)

[solegalli]

Hi @VascoSch92

This is looking really good. The tests are great.

Regarding the logic, I think we can speed this up if we store the range instead of min max,, and if we use dictionaries instead of dataframes. Could you give that a go?

Feel free to start working on the dosctrings and on adding a user guide :)

[VascoSch92]

Hey @solegalli

I changed what requested.
Now,

• params_ is a dictionary Dict[str, pd.Series] with keys 'mean' and 'range'
• MeanNormalizationScaling -> MeanNormalizationScaler
• error message for constant columns is a little better :-)

[solegalli]

Hey @VascoSch92 really good work here. Thank you so much!

We need to add a few files to create the docs now. Would you be able to do that as well?

Thanks a lot for the hard work.

[solegalli]

Last but not least: we need to add the new module on the readme and on the frontpage of the documentation, which lives here: https://github.com/feature-engine/feature_engine/blob/main/docs/index.rst

Thank you!!

[VascoSch92]

Hey @solegalli

• added the documentation for the new scaler
• divided params_ into mean_ and range_. Now mean_ and range_ are pd.Series. We can also make them np.array if you want. Sklearn uses np.arrays, should we make the same?

[solegalli]

Hey @VascoSch92 thanks for the quick turnaround. The api docs look great.

[VascoSch92]

Hey @solegalli

I changed to dictionaries instead of pd.Series, and it works :-)

[solegalli]

Amazing! Thanks a lot!

We just need to add a description / demo to the user guide folder in the docs and we are good to go then :)

[VascoSch92]

let me give a look ;-)

[VascoSch92]

@solegalli quick question: in general feature engine use scaling transformers from sklearn. Should we also include examples of these transformers?

[VascoSch92]

Hey @solegalli

I updated the docs with a demo.

Let me know what do you think. It is just a first version :-)

[VascoSch92]

Hey @solegalli :-) did you have time to look at the latest changes?

[solegalli]

Pending acceptance of suggested changes: VascoSch92#1