Hello, in the fit_transform diagram or lecture on pre-processing, I wonder why the target is not transformed before being injected in the model while it is the case in the next diagram.
Transformers do not process the target. The diagrams fit, transform, and fit_transform do not show the target at any point. It only shows the data (data or training data depending on if the method fit is involved).
Sorry about my mistake.
I meant the “pipeline fit diagram” which use an “unormalized” target contrary to the “pipeline predict” diagram, where the target is normalized before being processed.
pipeline.fit(X, y) does not transform the target and pipeline.predict(X) does not require the target y. The job of predict is to provide a predicted target. Potentially, you can compare this predicted target with the true target (if available) to compute a score.
Hello,
It seems I need some rest 
Thank you about your aswer (even if my issue was stupid).
It is easy to overlooked diagram
@Thierry71 don’t worry there is no stupid question. On the contrary in my opinion, it points to misunderstandings that probably many people have but they are too shy to ask about …