Fix/inverse transform for static covariate with single category across series (#2710)

* fix: bug when several static covariates have only one category

* feat: added corresponding tests

* update changelog

* update changelog

---------

Co-authored-by: dennisbader <[email protected]>

Author: madtoinou

CHANGELOG.md

@@ -21,6 +21,7 @@ but cannot always guarantee backwards compatibility. Changes that may **break co
 **Fixed**
 
 - 🔴 / 🟢 Fixed a bug which raised an error when loading torch models that were saved with Darts versions < 0.33.0. This is a breaking change and models saved with version 0.33.0 will not be loadable anymore. [#2692](https://github.com/unit8co/darts/pull/2692) by [Dennis Bader](https://github.com/dennisbader).
+- Fixed a bug in `StaticCovariatesTransformer` which raised an error when trying to inverse transform one-hot encoded categorical static covariates with identical values across time-series. Each categorical static covariates is now referred to by `{covariate_name}_{category_name}`, regardless of the number of categories. [#2710](https://github.com/unit8co/darts/pull/2710) by [Antoine Madrona](https://github.com/madtoinou)
 - Fixed a bug in `13-TFT-examples.ipynb` where two calls to `TimeSeries.from_series()` were not providing `series` but `pd.Index`. The method calls were changed to `TimeSeries.from_values()`. [#2719](https://github.com/unit8co/darts/pull/2719) by [Jules Authier](https://github.com/authierj)
 
 **Dependencies**

darts/dataprocessing/transformers/static_covariates_transformer.py

@@ -300,12 +300,8 @@ def _create_category_mappings(
                 for col, categories in zip(cols_cat, transformer_cat.categories_):
                     col_map_cat_i = []
                     for cat in categories:
-                        col_map_cat_i.append(cat)
-                        if len(categories) > 1:
-                            cat_col_name = str(col) + "_" + str(cat)
-                            inv_col_map_cat[cat_col_name] = [col]
-                        else:
-                            inv_col_map_cat[cat] = [col]
+                        col_map_cat_i.append(str(col) + "_" + str(cat))
+                        inv_col_map_cat[str(col) + "_" + str(cat)] = [col]
                     col_map_cat[col] = col_map_cat_i
         # If we don't have any categorical static covariates, don't need to generate mapping:
         else:
@@ -393,16 +389,6 @@ def _transform_static_covs(
             series, mask_num, mask_cat
         )
 
-        # Transform static covs:
-        tr_out_num, tr_out_cat = None, None
-        if mask_num.any():
-            tr_out_num = getattr(transformer_num, method)(vals_num)
-        if mask_cat.any():
-            tr_out_cat = getattr(transformer_cat, method)(vals_cat)
-            # sparse one hot encoding to dense array
-            if isinstance(tr_out_cat, csr_matrix):
-                tr_out_cat = tr_out_cat.toarray()
-
         # quick check if everything is in order
         n_vals_cat_cols = 0 if vals_cat is None else vals_cat.shape[1]
         if (method == "inverse_transform") and (n_vals_cat_cols != n_cat_cols):
@@ -413,6 +399,16 @@ def _transform_static_covs(
                 logger,
             )
 
+        # Transform static covs:
+        tr_out_num, tr_out_cat = None, None
+        if mask_num.any():
+            tr_out_num = getattr(transformer_num, method)(vals_num)
+        if mask_cat.any():
+            tr_out_cat = getattr(transformer_cat, method)(vals_cat)
+            # sparse one hot encoding to dense array
+            if isinstance(tr_out_cat, csr_matrix):
+                tr_out_cat = tr_out_cat.toarray()
+
         series = StaticCovariatesTransformer._add_back_static_covs(
             series, tr_out_num, tr_out_cat, mask_num, mask_cat, col_map_cat
         )
@@ -458,8 +454,6 @@ def _add_back_static_covs(
             elif is_cat:  # categorical transformed column
                 # covers one to one feature map (ordinal/label encoding) and one to multi feature (one hot encoding)
                 for col_name in col_map_cat[col]:
-                    if len(col_map_cat[col]) > 1:
-                        col_name = str(col) + "_" + str(col_name)
                     if col_name not in static_cov_columns:
                         data[col_name] = vals_cat[:, idx_cat]
                         static_cov_columns.append(col_name)

darts/tests/dataprocessing/transformers/test_static_covariates_transformer.py

@@ -184,6 +184,32 @@ def test_scaling_multi_series(self):
             series_recovered_multi[1].static_covariates
         )
 
+    def test_zero_cardinality_multi_series(self):
+        """Check that inverse-transform works as expected when OneHotEncoder is used on several series with
+        identical static covariates categories and values.
+        """
+        ts1 = self.series.with_static_covariates(
+            pd.Series({
+                "cov_a": "foo",
+                "cov_b": "foo",
+                "cov_c": "foo",
+            })
+        )
+        ts2 = self.series.with_static_covariates(
+            pd.Series({
+                "cov_a": "foo",
+                "cov_b": "foo",
+                "cov_c": "bar",
+            })
+        )
+
+        transformer = StaticCovariatesTransformer(transformer_cat=OneHotEncoder())
+        transformer.fit([ts1, ts2])
+        ts1_enc, ts2_enc = transformer.transform([ts1, ts2])
+        ts1_inv, ts2_inv = transformer.inverse_transform([ts1_enc, ts2_enc])
+        pd.testing.assert_frame_equal(ts1_inv.static_covariates, ts1.static_covariates)
+        pd.testing.assert_frame_equal(ts2_inv.static_covariates, ts2.static_covariates)
+
     def helper_test_scaling(self, series, scaler, test_values):
         series_tr = scaler.fit_transform(series)
         assert all([