[BUG] AttributeError: 'GroupedPredictor' object has no attribute 'predict_proba'

I am not sure if it’s me that I am not using correctly the API.

When doing a classification problem and using a GroupedPredictor to estimate probabilities I can’t get the probabilities

from sklego.meta import GroupedPredictor
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklego.datasets import load_chicken

df = load_chicken(as_frame=True)

# Create a binary target
X = df.drop(columns='weight')
y = np.where(df.weight>df.weight.mean(),1,0)

mod = GroupedPredictor(LogisticRegression(), groups=["diet"])

mod.fit(X,y)

mod.predict_proba(X)

Traceback (most recent call last): File "bug.py", line 25, in <module> mod.predict_proba(X) AttributeError: 'GroupedPredictor' object has no attribute 'predict_proba'

[skylarbpayne]

This can be closed, right?

[cmougan]

It might be the case that since our GroupedPredictor needs to support both regression and classification that we didn’t implement predict_proba. Can you confirm that predict does work?

I do confirm that in this exact code, if you change mod.predict_proba, by mod. predict you get an output