Fix sparse-lagged transformation for multi-ts

cases/multi_ts_level_forecasting.py

@@ -67,9 +67,7 @@ def run_multi_ts_forecast(forecast_length, is_multi_ts):
                   max_arity=4,
                   cv_folds=None,
                   validation_blocks=None,
-                  initial_assumption=init_pipeline,
-                  available_operations=['lagged', 'smoothing', 'diff_filter', 'gaussian_filter',
-                                        'ridge', 'lasso', 'linear', 'cut']
+                  initial_assumption=init_pipeline
                   )
     # fit model
     pipeline = model.fit(train_data)

fedot/core/operations/evaluation/operation_implementations/data_operations/ts_transformations.py

@@ -272,13 +272,7 @@ def _current_target_for_each_ts(self, current_ts_id, target):
     def _apply_transformation_for_predict(self, input_data: InputData):
         """Apply lagged transformation for every column (time series) in the dataset
         """
-
-        if self.sparse_transform:
-            self.log.debug(f'Sparse lagged transformation applied. If new data were used. Call fit method')
-            transformed_cols = self._update_features_for_sparse(input_data)
-            # Take last row in the lagged table and reshape into array with 1 row and n columns
-            self.features_columns = transformed_cols[-1].reshape(1, -1)
-            return self.features_columns
+        old_idx = copy(input_data.idx)
 
         if len(input_data.features.shape) > 1:
             # Multivariate time series
@@ -294,6 +288,12 @@ def _apply_transformation_for_predict(self, input_data: InputData):
             else:
                 current_ts = np.ravel(input_data.features[:, current_ts_id])
 
+            if self.sparse_transform:
+                self.log.debug('Sparse lagged transformation applied. If new data were used. Call fit method')
+                transformed_cols = self._update_features_for_sparse(current_ts, old_idx)
+                # Take last row in the lagged table and reshape into array with 1 row and n columns
+                current_ts = transformed_cols[-1].reshape(1, -1)
+
             # Take last window_size elements for current ts
             last_part_of_ts = current_ts[-self.window_size:].reshape(1, -1)
             if current_ts_id == 0:
@@ -320,13 +320,13 @@ def stack_by_type_predict(self, input_data, all_features, part_to_add):
             all_features = np.hstack((all_features, part_to_add))
         return all_features
 
-    def _update_features_for_sparse(self, input_data: InputData):
+    def _update_features_for_sparse(self, time_series: np.array, idx: np.array):
         """Make sparse matrix which will be used during forecasting
         """
 
         # Prepare features for training
-        new_idx, transformed_cols = ts_to_table(idx=input_data.idx,
-                                                time_series=input_data.features,
+        new_idx, transformed_cols = ts_to_table(idx=idx,
+                                                time_series=time_series,
                                                 window_size=self.window_size,
                                                 is_lag=True)
         # Sparsing