Better terms for predictor, regressor, classifier throughout the course

[lesteve]

This is probably too much jargon, also there is a source for confusion because regressors can also mean features used in a regression (similar thing for predictors) as mentioned in the scikit-learn glossary for example: https://scikit-learn.org/stable/glossary.html#term-predictor

In statistics, "predictors" refers to features.

The terms we agreed on:

• predictor -> predictive model
• regressor -> regression model
• classifier -> classification model

[lesteve]

We probably want to remove or update the corresponding entries in the glossary ...

[ArturoAmorQ]

We probably want to remove or update the corresponding entries in the glossary ...

Just a bit awkward that "Classifier" and "Regressor" would still be embedded in KNeighborsClassifier and DecisionTreeRegressor but I cannot find any simple work-around in this case.

[lesteve]

Good point we may want to keep the regressor and classifier entries in the glossary then ...

Another point, probably not that important, but my feeling was also that "regressor"/"classifier" was quite often used to mean a scikit-learn object as in the glossary definition and that "regression model"/"classification model" has a more general meaning.

[ArturoAmorQ]

One option is to keep both terms in the glossary and add a small explanation, for instance

### regression model

A regression model is a [predictive model](#predictive model) in a [regression](#regression)
setting.

In scikit-learn, `DecisionTreeRegressor` and `Ridge` are regression model classes

### regressor

The term "regressor" is quite often used to mean a scikit-learn object implementing a [regression model] (#regression model). 

Note:  "regressors" can also mean features used in a regression 

Or even just keeping the latter definition may suffice. What do you think?

[GaelVaroquaux]

What do you think?

I like your suggestion above!

[lesteve]

The weak consensus from Tuesday meeting is that it was not worth it, too much effort for too little impact.