Don't rely type_of_target, it is pretty buggy #293
Description
import sklearn.datasets
from sklearn.metrics import classification as classification_metrics
dataset = sklearn.datasets.load_linnerud()
classification_metrics.type_of_target(dataset.target)Returns multiclass-multioutput instead continuous-multioutput.
Similarly the following return multiclass, but it should regression.
import sklearn.datasets
from sklearn.metrics import classification as classification_metrics
dataset = sklearn.datasets.load_diabetes()
classification_metrics.type_of_target(dataset.target)I think AutoSklearnClassifier and AutoSklearnRegressor should just be replaced with one, AutoModel or something which simply tries to run both classifier and regressor and see what works better.