BUG fix behaviour in confusion_matrix with with empty array-like as input

[parsons-kyle-89]

Reference Issues/PRs

Fixes #16432

What does this implement/fix? Explain your changes.

Currently on master confusion_matrix raises a ValueError when y_true and y_pred are empty and labels is not None.

from sklearn.metrics import confusion_matrix

confusion_matrix([], [])
# returns a 0x0 numpy array of dtype int64

confusion_matrix([], [], labels=[True, False])
# raises ValueError

With the change in this PR, the second case will return an nxn numpy array of zeros. This is correct with respect to the definition of confusion matrix given in the documentation.

Any other comments?

This change should be especially useful to users using confusion_matrix in automated reporting. Automated reports will sometimes (correctly) be run on datasets of size 0. In this case allowing confusion_matrix to correctly return an nxn matrix of zeros rather than raising will eliminate the need for the user to write a special case.

[glemaitre]

Could you add the following:

• an entry to the change log at doc/whats_new/v0.23.rst under bug fixes. Like the other entries there, please reference this pull request with :pr: and credit yourself (and other contributors if applicable) with :user:
• a similar test to check that plot_confusion_matrix is giving the appropriate result as well. The file is located in sklearn/metrics/_plot/tests/test_plot_confusion_matrix.py.

[parsons-kyle-89]

* a similar test to check that `plot_confusion_matrix` is giving the appropriate result as well. The file is located in `sklearn/metrics/_plot/tests/test_plot_confusion_matrix.py`.

I wasn't quite sure what form such a test should have, but I copied the form of another test of text color. Let me know if you think there should be a more thorough test here.

[glemaitre]

last changes otherwise LGTM.

[glemaitre]

I pushed my last comments and split one of the test which focus only on errors raised in confusion_matrix. I will merge if CIs turn green.

[glemaitre]

@parsons-kyle-89 Thank for raising and contributing

[parsons-kyle-89]

@parsons-kyle-89 Thank for raising and contributing

👍 Glad to contribute. Thanks for being responsive and getting it over the finish line.