[enhancement] add predict_proba for classifiers about autofeat HOT 11 CLOSED

@mglowacki100 this is what it looks like for me:

from autofeat.

wow, what a horrible UX. sorry, now you should see it

from autofeat.

Thanks so much, I merged the changes and will release a new version next weekend (since I also wanted to make sure the code works with the new sklearn versions and everything)

from autofeat.

@cod3licious Thanks, it works correctly :)

from autofeat.

Great idea! Unfortunately, I don't know when I'll get to that, but pull requests are always welcome!

from autofeat.

Hi
@cod3licious cod3 as far I see and after running local test it seems enough to add method predict_proba as minimally modified predict method to class AutoFeatModel(BaseEstimator)

    def predict_proba(self, X):
        """
        Inputs:
            - X: pandas dataframe or numpy array with original features (n_datapoints x n_features)
        Returns:
            - y_pred: predicted targets probs return by prediction_model.predict_proba()
        """
        check_is_fitted(self, ["prediction_model_"])
        # store column names as they'll be lost in the other check
        cols = [str(c) for c in X.columns] if isinstance(X, pd.DataFrame) else []
        # check input variables
        X = check_array(X, dtype=None)
        if not cols:
            cols = ["x%03i" % i for i in range(X.shape[1])]
        # transform X into a dataframe (again)
        df = pd.DataFrame(X, columns=cols)
        # do we need to call transform?
        if not list(df.columns) == self.all_columns_:
            temp = self.always_return_numpy
            self.always_return_numpy = False
            df = self.transform(df)
            self.always_return_numpy = temp
        return self.prediction_model_.predict_proba(df[self.good_cols_].to_numpy())

I'll try to make proper PR in the coming week.

from autofeat.

Here is PR:
#38

from autofeat.

@mglowacki100 I've commented on your PR, it would be great if you could change the code accordingly!

from autofeat.

Hi @cod3licious,
Maybe, I miss something obvious but I don't see any comments/reviews for this PR :(

from autofeat.

@cod3licious I think 'Pending' status of review is misleading (https://github.com/orgs/community/discussions/10369)
On my side I see it like this:

from autofeat.

@mglowacki100 you should now be able to install version 2.1.0 with the predict_proba function using pip install --upgrade autofeat - please let me know if there are any problems :)

from autofeat.