On transforming after fit_transform about featurewiz

[AutoViML]

Hi @asimovidick 👍

If you get the features.features object, you already have selected the best features from your dataset that are likely to have a strong signal. You can then use this "filter" (as you call it) to later transform those selected features either by one-hot encoding or by label encoding or by using any other feature transformation you want.

You must use the features.features filter only once to select the strongest features in your dataset.

Don't try to use it repeatedly (like a sklearn transformer) to select features since you don't need to select features every time you run cross-validation or an inference in an MLOps pipeline. That's why the transform function in features object was set to be a filter. It was not meant to be used during inferencing or cross-validation.

I hope this explanation makes sense.
AutoViML

[asimovidick]

I got it! Thanks, and it make sense, featurewiz selects features, it is not a simple transformer.

Thank you again all you guys
Cheers