DataFrameMapper.inverse_transform() for simple transformations

sklearn_pandas/dataframe_mapper.py

@@ -110,6 +110,7 @@ def __init__(self, features, default=False, sparse=False, df_out=False,
         self.df_out = df_out
         self.input_df = input_df
         self.transformed_names_ = []
+        self.transformed_cols_ = []
 
         if (df_out and (sparse or default)):
             raise ValueError("Can not use df_out with sparse or default")
@@ -268,6 +269,7 @@ def transform(self, X):
         """
         extracted = []
         self.transformed_names_ = []
+        self.transformed_cols_ = []
         for columns, transformers, options in self.built_features:
             input_df = options.get('input_df', self.input_df)
             # columns could be a string or list of
@@ -283,6 +285,10 @@ def transform(self, X):
             self.transformed_names_ += self.get_names(
                 columns, transformers, Xt, alias)
 
+            self.transformed_cols_ += [
+              (columns, transformers,
+               self.get_names(columns, transformers, Xt, alias))]
+
         # handle features not explicitly selected
         if self.built_default is not False:
             unsel_cols = self._unselected_columns(X)
@@ -328,3 +334,34 @@ def transform(self, X):
                                 index=index)
         else:
             return stacked
+
+    def inverse_transform(self, X):
+        """
+        Inverse transform the given data. Assumes that fit has already been
+        called.
+
+        X       the data to inverse transform
+        """
+
+        X_inv = pd.DataFrame()
+        # We will populate the inverse transformed dataframe column by column
+
+        # Let's keep track of the column we've processed
+        prev_col = 0
+        for columns, transformers, transformed_cols in self.transformed_cols_:
+            # Determine the column number of the last column in X
+            # corresponding to the original column we're computing
+            last_col = prev_col + len(transformed_cols)
+
+            # Inverse transform the columns in X for the current transformer
+            col_inv = pd.DataFrame(transformers.inverse_transform(
+                                      X[:, prev_col:last_col]),
+                                   columns=[columns])
+
+            # Append the inverse transformed column to the output data frame
+            X_inv = pd.concat([X_inv, col_inv], axis=1)
+
+            # For the next iteration, update the last column processed
+            prev_col = last_col
+
+        return X_inv

tests/test_dataframe_mapper.py

@@ -19,7 +19,7 @@
 from sklearn.feature_extraction.text import CountVectorizer
 from sklearn.feature_extraction import DictVectorizer
 from sklearn.preprocessing import (
-    Imputer, StandardScaler, OneHotEncoder, LabelBinarizer)
+    Imputer, StandardScaler, OneHotEncoder, LabelBinarizer, LabelEncoder)
 from sklearn.feature_selection import SelectKBest, chi2
 from sklearn.base import BaseEstimator, TransformerMixin
 import sklearn.decomposition
@@ -829,3 +829,34 @@ def test_direct_cross_validation(iris_dataframe):
     scores = sklearn_cv_score(pipeline, data, labels)
     assert scores.mean() > 0.96
     assert (scores.std() * 2) < 0.04
+
+
+def test_inverse_transform_simple():
+    df = pd.DataFrame({'colA': list('ynyyn'), 'colB': list('abcab')})
+    mapper = DataFrameMapper([
+        ('colA', LabelEncoder()),
+        ('colB', LabelEncoder()),
+    ])
+
+    transformed = mapper.fit_transform(df)
+    restored = mapper.inverse_transform(transformed)
+
+    assert isinstance(restored, pd.DataFrame)
+    assert restored.equals(df)
+
+
+def test_inverse_transform_multicolumn():
+    df = pd.DataFrame({'colA': list('ynyyn'),
+                       'colB': list('abcab'),
+                       'colC': list('sttts')})
+    mapper = DataFrameMapper([
+        ('colA', LabelEncoder()),
+        ('colB', LabelBinarizer()),
+        ('colC', LabelEncoder()),
+    ])
+
+    transformed = mapper.fit_transform(df)
+    restored = mapper.inverse_transform(transformed)
+
+    assert isinstance(restored, pd.DataFrame)
+    assert restored.equals(df)