This a general question regarding the API of as_frame and return_X_y with our loader/fetcher. We have an inconsistent behaviour which could be solved straight away.

fetch_opeml introduced as_frame exposing frame attribute in the Bunch which is a Pandas DataFrame. In conjunction with return_X_y=True, we exposed data and target. data will always be a DataFrame while target is supposed to be a DataFrame or a Series depending on the number of columns in the target.

X, y = fetch_openml('iris', as_frame=True, return_X_y=True)
y

0         Iris-setosa
1         Iris-setosa
2         Iris-setosa
3         Iris-setosa
4         Iris-setosa
            ...      
145    Iris-virginica
146    Iris-virginica
147    Iris-virginica
148    Iris-virginica
149    Iris-virginica
Name: class, Length: 150, dtype: category
Categories (3, object): [Iris-setosa, Iris-versicolor, Iris-virginica]

In #15950, we introduce as_frame to fetch_california_housing. The API is the same apart of the target output. The target is a DataFrame even with a single column.

X, y = fetch_california_housing(return_X_y=True, as_frame=True)
y

       MedHouseVal
0            4.526
1            3.585
2            3.521
3            3.413
4            3.422
...            ...
20635        0.781
20636        0.771
20637        0.923
20638        0.847
20639        0.894

[20640 rows x 1 columns]

So my question is: what type of target do we want when the target is 1D?