is it a bug?

[bringtree]

Description

an err output, In my train_output, it does not exist [0,0]

Steps/Code to Reproduce

from sklearn import tree

X = [[1, 1], [1, 2], [1, 3], [1, 4], [2, 1], [2, 2], [2, 3], [2, 4]]
Y = [[0, 1], [0, 1], [1, 0], [1, 0], [1, 0], [1, 0], [1, 0], [1, 0]]
# Y = [[0], [0], [1], [1], [1], [1], [1], [1]]

clf = tree.DecisionTreeClassifier(criterion='entropy', max_depth=1)
clf.fit(X, Y)
print(clf.predict([[1, 1], [1, 2], [1, 3], [1, 4], [2, 1], [2, 2], [2, 3], [2, 4]]))

Expected Results

[[ 0. 1.]
[ 0. 1.]
[ 0. 1.]
[ 0. 1.]
[ 1. 0.]
[ 1. 0.]
[ 1. 0.]
[ 1. 0.]]

Actual Results

[[ 0. 0.]
[ 0. 0.]
[ 0. 0.]
[ 0. 0.]
[ 1. 0.]
[ 1. 0.]
[ 1. 0.]
[ 1. 0.]]

Versions

Darwin-16.7.0-x86_64-i386-64bit
Python 3.6.1 (v3.6.1:69c0db5050, Mar 21 2017, 01:21:04)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
NumPy 1.13.3
SciPy 0.19.1
Scikit-Learn 0.19.0

[amueller]

if only we had #9251....

[amueller]

No. It's a tie-breaking mechanism. I think we actually always break ties to the smallest class, not randomly. I'm not sure if that is ideal, but either way, this is not bug.

I don't understand your expected result btw.

[bringtree]

@amueller Thanks. But the result [0,0] does not exist in my train output.
When i don't use the one-hot encoder, this situation will not appear.

[jnothman]

you shouldn't be one-hot encoding targets for multiclass classification

[amueller]

encoding the targets as one-hot indicates to the estimator that you want it treated as a multi-label problem.

[bringtree]

oh, thanks