commit-0
diff --git a/‎doc/whats_new/upcoming_changes/sklearn.metrics/29213.enhancement.rst
Lines changed: 3 additions & 0 deletions b/‎doc/whats_new/upcoming_changes/sklearn.metrics/29213.enhancement.rst
Lines changed: 3 additions & 0 deletions
diff --git a/‎sklearn/metrics/_classification.py
Lines changed: 24 additions & 1 deletion b/‎sklearn/metrics/_classification.py
Lines changed: 24 additions & 1 deletion
diff --git a/‎sklearn/metrics/tests/test_classification.py
Lines changed: 12 additions & 0 deletions b/‎sklearn/metrics/tests/test_classification.py
Lines changed: 12 additions & 0 deletions
@@ -0,0 +1,3 @@
+- :func:`sklearn.metrics.accuracy_score` now includes a `zero_division`
+  parameter to raise a warning when  `y_true` and `y_pred` are empty.
+  By :user:`Jaimin Chauhan <jaimin020>`.
@@ -152,10 +152,16 @@ def _check_targets(y_true, y_pred):
         "y_pred": ["array-like", "sparse matrix"],
         "normalize": ["boolean"],
         "sample_weight": ["array-like", None],
+        "zero_division": [
+            Options(Real, {0.0, 1.0, np.nan}),
+            StrOptions({"warn"}),
+        ],
     },
     prefer_skip_nested_validation=True,
 )
-def accuracy_score(y_true, y_pred, *, normalize=True, sample_weight=None):
+def accuracy_score(
+    y_true, y_pred, *, normalize=True, sample_weight=None, zero_division="warn"
+):
     """Accuracy classification score.
 
     In multilabel classification, this function computes subset accuracy:
@@ -179,6 +185,13 @@ def accuracy_score(y_true, y_pred, *, normalize=True, sample_weight=None):
     sample_weight : array-like of shape (n_samples,), default=None
         Sample weights.
 
+    zero_division : {"warn", 0.0, 1.0, np.nan}, default="warn"
+        Sets the value to return when there is a zero division,
+        e.g. when `y_true` and `y_pred` are empty.
+        If set to "warn", returns 0.0 input, but a warning is also raised.
+
+        versionadded:: 1.6
+
     Returns
     -------
     score : float or int
@@ -220,6 +233,16 @@ def accuracy_score(y_true, y_pred, *, normalize=True, sample_weight=None):
     y_true, y_pred = attach_unique(y_true, y_pred)
     y_type, y_true, y_pred = _check_targets(y_true, y_pred)
     check_consistent_length(y_true, y_pred, sample_weight)
+
+    
8000
if _num_samples(y_true) == 0:
+        if zero_division == "warn":
+            msg = (
+                "accuracy() is ill-defined and set to 0.0. Use the `zero_division` "
+                "param to control this behavior."
+            )
+            warnings.warn(msg, UndefinedMetricWarning)
+        return _check_zero_division(zero_division)
+
     if y_type.startswith("multilabel"):
         if _is_numpy_namespace(xp):
             differing_labels = count_nonzero(y_true - y_pred, axis=1)
 
@@ -809,13 +809,19 @@ def test_matthews_corrcoef_nan():
         partial(fbeta_score, beta=1),
         precision_score,
         recall_score,
+        accuracy_score,
         partial(cohen_kappa_score, labels=[0, 1]),
     ],
 )
 def test_zero_division_nan_no_warning(metric, y_true, y_pred, zero_division):
     """Check the behaviour of `zero_division` when setting to 0, 1 or np.nan.
     No warnings should be raised.
     """
+    if metric is accuracy_score and len(y_true):
+        pytest.skip(
+            reason="zero_division is only used with empty y_true/y_pred for accuracy"
+        )
+
     with warnings.catch_warnings():
         warnings.simplefilter("error")
         result = metric(y_true, y_pred, zero_division=zero_division)
@@ -834,13 +840,19 @@ def test_zero_division_nan_no_warning(metric, y_true, y_pred, zero_division):
         partial(fbeta_score, beta=1),
         precision_score,
         recall_score,
+        accuracy_score,
         cohen_kappa_score,
     ],
 )
 def test_zero_division_nan_warning(metric, y_true, y_pred):
     """Check the behaviour of `zero_division` when setting to "w
6F50
arn".
     A `UndefinedMetricWarning` should be raised.
     """
+    if metric is accuracy_score and len(y_true):
+        pytest.skip(
+            reason="zero_division is only used with empty y_true/y_pred for accuracy"
+        )
+
     with pytest.warns(UndefinedMetricWarning):
         result = metric(y_true, y_pred, zero_division="warn")
     assert result == 0.0
Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,3 @@`
	`1`	+- :func:`sklearn.metrics.accuracy_score` now includes a `zero_division`
	`2`	+ parameter to raise a warning when `y_true` and `y_pred` are empty.
	`3`	+ By :user:`Jaimin Chauhan <jaimin020>`.