fix: handle category dtype for target column in _build_dataframe

supervised/base_automl.py

@@ -673,7 +673,13 @@ def _handle_drastic_imbalance(
                         ] = new_sensitive_features.loc[j]
 
     def _save_data_info(self, X, y, sample_weight=None, sensitive_features=None):
-        target_is_numeric = pd.api.types.is_numeric_dtype(y)
+        # Resolve category dtype before numeric check so that targets like
+        # pd.Series([0, 1, 1]).astype('category') are correctly identified
+        # as numeric (the categories themselves are numeric).
+        if isinstance(y, pd.Series) and isinstance(y.dtype, pd.CategoricalDtype):
+            target_is_numeric = pd.api.types.is_numeric_dtype(y.cat.categories)
+        else:
+            target_is_numeric = pd.api.types.is_numeric_dtype(y)
         if self._ml_task == MULTICLASS_CLASSIFICATION:
             y = y.astype(str)
 
@@ -792,6 +798,19 @@ def _build_dataframe(self, X, y=None, sample_weight=None, sensitive_features=Non
             y = check_array(y, ensure_2d=False)
             y = pd.Series(np.array(y), name="target")
 
+        # If y is a pd.Series with category dtype (e.g. user called
+        # df[col].astype('category') on the target column), convert it to
+        # the underlying dtype so that downstream code (LightGBM, sklearn,
+        # dtype checks, etc.) can handle it correctly.
+        if isinstance(y, pd.Series) and isinstance(y.dtype, pd.CategoricalDtype):
+            if pd.api.types.is_numeric_dtype(y.cat.categories):
+                # Numeric categories (e.g. 0, 1, 2 stored as category) —
+                # restore to the actual numeric dtype.
+                y = y.astype(y.cat.categories.dtype)
+            else:
+                # String/object categories — convert to plain object dtype.
+                y = y.astype(object)
+
         if sample_weight is not None:
             if isinstance(sample_weight, np.ndarray):
                 sample_weight = check_array(sample_weight, ensure_2d=False)

tests/tests_automl/test_data_types.py

@@ -45,3 +45,69 @@ def test_encoding_strange_characters(self):
             start_random_models=1,
         )
         automl.fit(X, y)
+
+    def test_numeric_category_target(self):
+        """Regression test: target stored as numeric category dtype should not
+        raise 'ValueError: pandas dtypes must be int, float or bool'.
+        See: https://github.com/mljar/mljar-supervised/issues/XXX
+        """
+        X = np.random.rand(self.rows, 3)
+        X = pd.DataFrame(X, columns=[f"f{i}" for i in range(3)])
+        # Binary target with integer values stored as category dtype
+        y = pd.Series(np.random.randint(0, 2, self.rows)).astype("category")
+
+        automl = AutoML(
+            results_path=self.automl_dir,
+            total_time_limit=1,
+            algorithms=["Baseline"],
+            train_ensemble=False,
+            explain_level=0,
+            start_random_models=1,
+        )
+        automl.fit(X, y)
+
+    def test_numeric_category_features_and_target(self):
+        """Regression test: both X feature columns and y target stored as
+        numeric category dtype must be handled without errors.
+        """
+        X = pd.DataFrame(
+            {
+                "f0": np.random.rand(self.rows),
+                # Numeric-valued categorical feature — the exact scenario from the issue
+                "f1": pd.Series(
+                    np.random.randint(1, 5, self.rows)
+                ).astype("category"),
+                "f2": pd.Series(
+                    np.random.randint(0, 3, self.rows)
+                ).astype("category"),
+            }
+        )
+        y = pd.Series(np.random.randint(0, 2, self.rows)).astype("category")
+
+        automl = AutoML(
+            results_path=self.automl_dir,
+            total_time_limit=1,
+            algorithms=["Baseline"],
+            train_ensemble=False,
+            explain_level=0,
+            start_random_models=1,
+        )
+        automl.fit(X, y)
+
+    def test_string_category_target(self):
+        """Regression test: target with string-valued category dtype is handled."""
+        X = np.random.rand(self.rows, 3)
+        X = pd.DataFrame(X, columns=[f"f{i}" for i in range(3)])
+        y = pd.Series(
+            np.random.choice(["cat", "dog"], size=self.rows)
+        ).astype("category")
+
+        automl = AutoML(
+            results_path=self.automl_dir,
+            total_time_limit=1,
+            algorithms=["Baseline"],
+            train_ensemble=False,
+            explain_level=0,
+            start_random_models=1,
+        )
+        automl.fit(X, y)