Fix bug in rfe when only 1 feature left (#639)

* add problematic test

* fixes bug

* fix code style

Author: solegalli

feature_engine/selection/recursive_feature_elimination.py

@@ -169,6 +169,12 @@ def fit(self, X: pd.DataFrame, y: pd.Series):
         # remember that feature_importances_ is ordered already
         for feature in list(self.feature_importances_.index):
 
+            # if there is only 1 feature left
+            if X_tmp.shape[1] == 1:
+                self.performance_drifts_[feature] = 0
+                _selected_features.append(feature)
+                break
+
             # remove feature and train new model
             model_tmp = cross_validate(
                 self.estimator,
@@ -196,11 +202,6 @@ def fit(self, X: pd.DataFrame, y: pd.Series):
                 # remove feature and adjust initial performance
                 X_tmp = X_tmp.drop(columns=feature)
 
-                if X_tmp.empty is True:
-                    raise ValueError(
-                        "All features have been removed. Try reducing the threshold."
-                    )
-
                 baseline_model = cross_validate(
                     self.estimator,
                     X_tmp,

tests/test_selection/test_recursive_feature_elimination.py

@@ -1,7 +1,7 @@
 import pandas as pd
 import pytest
 from sklearn.ensemble import RandomForestClassifier
-from sklearn.linear_model import Lasso, LogisticRegression
+from sklearn.linear_model import Lasso, LogisticRegression, LinearRegression
 from sklearn.tree import DecisionTreeRegressor
 
 from feature_engine.selection import RecursiveFeatureElimination
@@ -186,3 +186,25 @@ def test_regression(
 
     # test transform output
     pd.testing.assert_frame_equal(sel.transform(X), Xtransformed)
+
+
+def test_stops_when_only_one_feature_remains():
+    linear_model = LinearRegression()
+
+    # Feature x shows 100% correlation with target variable
+    # Feature x shows 0% correlation with target variable
+    # Target variable: y
+
+    df = pd.DataFrame(
+        {
+            "x": [0, 0, 0, 0, 0, 1, 1, 1, 1, 1],
+            "z": [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
+            "y": [0, 0, 0, 0, 0, 1, 1, 1, 1, 1],
+        }
+    )
+
+    transformer = RecursiveFeatureElimination(
+        estimator=linear_model, scoring="r2", cv=3
+    )
+    output = transformer.fit_transform(df[["x", "z"]], df["y"])
+    pd.testing.assert_frame_equal(output, df["x"].to_frame())