swev-id: scikit-learn__scikit-learn-25931: Fix IsolationForest feature names warning during fit when contamination != 'auto'

sklearn/ensemble/_iforest.py

@@ -345,7 +345,9 @@ def fit(self, X, y=None, sample_weight=None):
             return self
 
         # else, define offset_ wrt contamination parameter
-        self.offset_ = np.percentile(self.score_samples(X), 100.0 * self.contamination)
+        self.offset_ = np.percentile(
+            self._score_samples_no_validation(X), 100.0 * self.contamination
+        )
 
         return self
 
@@ -435,6 +437,14 @@ def score_samples(self, X):
         # Check data
         X = self._validate_data(X, accept_sparse="csr", dtype=np.float32, reset=False)
 
+        return self._score_samples_no_validation(X)
+
+    def _score_samples_no_validation(self, X):
+        if issparse(X):
+            X = X.tocsr()
+        else:
+            X = np.asarray(X, dtype=np.float32)
+
         # Take the opposite of the scores as bigger is better (here less
         # abnormal)
         return -self._compute_chunked_score_samples(X)

sklearn/ensemble/tests/test_iforest.py

@@ -10,6 +10,7 @@
 import warnings
 
 import numpy as np
+import pandas as pd
 
 from sklearn.utils._testing import assert_array_equal
 from sklearn.utils._testing import assert_array_almost_equal
@@ -33,6 +34,12 @@
 diabetes = load_diabetes()
 
 
+FEATURE_NAMES_WARNING = (
+    "X does not have valid feature names, but IsolationForest was fitted with "
+    "feature names"
+)
+
+
 def test_iforest(global_random_seed):
     """Check Isolation Forest for various parameter settings."""
     X_train = np.array([[0, 1], [1, 2]])
@@ -75,6 +82,51 @@ def test_iforest_sparse(global_random_seed):
             assert_array_equal(sparse_results, dense_results)
 
 
+def test_iforest_fit_dataframe_contamination_no_warning():
+    X = pd.DataFrame({"a": [-1.1, 0.3, 0.5, 100]})
+
+    iso = IsolationForest(random_state=0, contamination=0.05)
+
+    with warnings.catch_warnings():
+        warnings.filterwarnings(
+            "error", message=FEATURE_NAMES_WARNING, category=UserWarning
+        )
+        warnings.filterwarnings("ignore", category=DeprecationWarning)
+        iso.fit(X)
+
+
+def test_iforest_dataframe_then_ndarray_warns_on_score_and_predict():
+    X = pd.DataFrame({"a": [-1.1, 0.3, 0.5, 100]})
+
+    iso = IsolationForest(random_state=0, contamination=0.05)
+
+    with warnings.catch_warnings():
+        warnings.simplefilter("ignore", category=DeprecationWarning)
+        iso.fit(X)
+
+    X_array = X.to_numpy()
+
+    with pytest.warns(UserWarning, match=FEATURE_NAMES_WARNING):
+        iso.score_samples(X_array)
+
+    with pytest.warns(UserWarning, match=FEATURE_NAMES_WARNING):
+        iso.predict(X_array)
+
+
+def test_iforest_fit_dataframe_auto_no_warning_and_offset():
+    X = pd.DataFrame({"a": [-1.1, 0.3, 0.5, 100]})
+    iso = IsolationForest(random_state=0, contamination="auto")
+
+    with warnings.catch_warnings():
+        warnings.filterwarnings(
+            "error", message=FEATURE_NAMES_WARNING, category=UserWarning
+        )
+        warnings.filterwarnings("ignore", category=DeprecationWarning)
+        iso.fit(X)
+
+    assert iso.offset_ == -0.5
+
+
 def test_iforest_error():
     """Test that it gives proper exception on deficient input."""
     X = iris.data