feat: add ml.metrics.mean_absolute_error method (#1910)

Author: GarrettWu

bigframes/ml/metrics/__init__.py

@@ -18,6 +18,7 @@
     auc,
     confusion_matrix,
     f1_score,
+    mean_absolute_error,
     mean_squared_error,
     precision_score,
     r2_score,
@@ -36,6 +37,7 @@
     "confusion_matrix",
     "precision_score",
     "f1_score",
+    "mean_absolute_error",
     "mean_squared_error",
     "pairwise",
 ]

bigframes/ml/metrics/_metrics.py

@@ -344,3 +344,17 @@ def mean_squared_error(
 mean_squared_error.__doc__ = inspect.getdoc(
     vendored_metrics_regression.mean_squared_error
 )
+
+
+def mean_absolute_error(
+    y_true: Union[bpd.DataFrame, bpd.Series],
+    y_pred: Union[bpd.DataFrame, bpd.Series],
+) -> float:
+    y_true_series, y_pred_series = utils.batch_convert_to_series(y_true, y_pred)
+
+    return (y_pred_series - y_true_series).abs().sum() / len(y_true_series)
+
+
+mean_absolute_error.__doc__ = inspect.getdoc(
+    vendored_metrics_regression.mean_absolute_error
+)

tests/system/small/ml/test_metrics.py

@@ -818,3 +818,10 @@ def test_mean_squared_error(session: bigframes.Session):
     df = session.read_pandas(pd_df)
     mse = metrics.mean_squared_error(df["y_true"], df["y_pred"])
     assert mse == 0.375
+
+
+def test_mean_absolute_error(session: bigframes.Session):
+    pd_df = pd.DataFrame({"y_true": [3, -0.5, 2, 7], "y_pred": [2.5, 0.0, 2, 8]})
+    df = session.read_pandas(pd_df)
+    mse = metrics.mean_absolute_error(df["y_true"], df["y_pred"])
+    assert mse == 0.5

third_party/bigframes_vendored/sklearn/metrics/_regression.py

@@ -91,3 +91,30 @@ def mean_squared_error(y_true, y_pred) -> float:
         float: Mean squared error.
     """
     raise NotImplementedError(constants.ABSTRACT_METHOD_ERROR_MESSAGE)
+
+
+def mean_absolute_error(y_true, y_pred) -> float:
+    """Mean absolute error regression loss.
+
+    **Examples:**
+
+        >>> import bigframes.pandas as bpd
+        >>> import bigframes.ml.metrics
+        >>> bpd.options.display.progress_bar = None
+
+        >>> y_true = bpd.DataFrame([3, -0.5, 2, 7])
+        >>> y_pred = bpd.DataFrame([2.5, 0.0, 2, 8])
+        >>> mae = bigframes.ml.metrics.mean_absolute_error(y_true, y_pred)
+        >>> mae
+        np.float64(0.5)
+
+    Args:
+        y_true (Series or DataFrame of shape (n_samples,)):
+            Ground truth (correct) target values.
+        y_pred (Series or DataFrame of shape (n_samples,)):
+            Estimated target values.
+
+    Returns:
+        float: Mean absolute error.
+    """
+    raise NotImplementedError(constants.ABSTRACT_METHOD_ERROR_MESSAGE)