Error: Algorithm should either be a classifier to be used with response_method=predict_proba or the response_method should be 'predict'

[yiannis-gkoufas]

Hi!

Recently I started getting this error (or warning?) with binary classification.
Minimal code to reproduce the problem:

from supervised import AutoML

if __name__ == '__main__':
    from sklearn.datasets import make_classification
    X, y = make_classification(n_samples=100000, n_features=20, n_redundant=2)
    from sklearn.model_selection import train_test_split
    X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.3)
    automl = AutoML(eval_metric="accuracy")
    automl.fit(X_train, y_train)
    automl.report()

I get multiple messages:

DecisionTreeAlgorithm should either be a classifier to be used with response_method=predict_proba or the response_method should be 'predict'. Got a regressor with response_method=predict_proba instead.
Problem during computing permutation importance. Skipping ...

Is it a new bug or we can just ignore it?
Thank you!

[pplonski]

Hi @yiannis-gkoufas,

Thank you for reporting the issue and thanks for code to reproduce the issue. I need to look closer into it. Hard to say is it bug or can be ignored. Any help will be appreciated :)

[yiannis-gkoufas]

Sounds like an update of sklearn might have caused it?

[williamty]

i met this error too. it will appear for DecisionTreeAlgorithm/LightgbmAlgorithm/XgbAlgorithm/RandomForestAlgorithm/ExtraTreesAlgorithm, all of which are tree based algorithm. i have set explain_level to 2. i'm not sure if the error is affected by the explain level.

[pplonski]

I'm not sleeping - fixing ... Looks like after sklearn update the warning started to popup. To fix it I've added ClassifierMixin or RegressorMixin from sklearn.base for each algorithm.

I've seen few more warnings, I would like to fix them and then release a new version.