[MRG] DOC mention the multi-class scheme used

[glemaitre]

Reference Issue

Partially addressing #303

What does this implement/fix? Explain your changes.

Mentioning in the docstring which type of multi-class scheme is used.
We need to add more information in the User Guide which is coming.

Any other comments?

[codecov]

Codecov Report

Merging #311 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #311   +/-   ##
=======================================
  Coverage   98.33%   98.33%           
=======================================
  Files          68       68           
  Lines        3910     3910           
=======================================
  Hits         3845     3845           
  Misses         65       65

Impacted Files	Coverage Δ
...prototype_selection/neighbourhood_cleaning_rule.py	100% <ø> (ø)	⬆️
...sampling/prototype_generation/cluster_centroids.py	100% <ø> (ø)	⬆️
...prototype_selection/condensed_nearest_neighbour.py	100% <ø> (ø)	⬆️
...prototype_selection/instance_hardness_threshold.py	97.43% <ø> (ø)	⬆️
imblearn/over_sampling/adasyn.py	97.77% <ø> (ø)	⬆️
imblearn/combine/smote_tomek.py	100% <ø> (ø)	⬆️
.../under_sampling/prototype_selection/tomek_links.py	100% <ø> (ø)	⬆️
imblearn/ensemble/easy_ensemble.py	100% <ø> (ø)	⬆️
imblearn/over_sampling/random_over_sampler.py	100% <ø> (ø)	⬆️
imblearn/ensemble/balance_cascade.py	100% <ø> (ø)	⬆️
... and 5 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f695fb7...77b47e7. Read the comment docs.

[glemaitre]

@chkoar @massich Should not be a hard one to be checked.

[chkoar]

Isn't it one-vs-one with rest?

[glemaitre]

one-vs-one with rest? I am not even sure that we should call like that since this is not for classifier.
In short this is the class of interest vs the other classes (which is why I called one-vs-rest since it really look like https://en.wikipedia.org/wiki/Multiclass_classification#One-vs.-rest).

[chkoar]

Oh no. For instance, I have a majority and three minorities, right? I have to over-sample all the rest classes (against the majority) for a ratio of the majority. That's why I named it like that.

[glemaitre]

I have to over-sample all the rest classes.

Uhm. This is more tricky since it is decided by the ratio itself which will target the class.
Then for each class targeted this is usually one-vs-rest.

It seems pretty confusing :D We need to emphasize in the doc. We could redirect to the documentation in the docstring. I don't think that we can add so much more in the docstring itself.

WDYT?