restore index.year

VegeWaterDynamics · Nov 24, 2023 · dc39649 · dc39649
1 parent 298b8b8
commit dc39649
Showing 1 changed file with 9 additions and 13 deletions.
diff --git a/motrainer/jackknife.py b/motrainer/jackknife.py
@@ -118,35 +118,31 @@ def train(self,
                                                       normalize_method)
 
         # Data split
-        input_years = self.gpi_input.index.get_level_values('year')
-        output_years = self.gpi_output.index.get_level_values('year')
+        input_years = self.gpi_input.index.year
+        output_years = self.gpi_output.index.year
         logger.debug(
             'Spliting Trainning and validation data. Split year: {}.'.format(
                 self.val_split_year))
         jackknife_input = self.gpi_input[input_years < self.val_split_year]
         jackknife_output = self.gpi_output[output_years < self.val_split_year]
         vali_input = self.gpi_input[input_years >= self.val_split_year]
         vali_output = self.gpi_output[output_years >= self.val_split_year]
-        year_list = jackknife_input.index.get_level_values('year').unique()
+        year_list = jackknife_input.index.year.unique()
 
         # Jackknife in time
         loo = LeaveOneOut()
         best_perf_sum = None
         for train_index, test_index in loo.split(year_list):
             this_year = test_index[0] + year_list[0]
 
-            input_years = jackknife_input.index.get_level_values('year')
-            output_years = jackknife_output.index.get_level_values('year')
+            input_years = jackknife_input.index.year
+            output_years = jackknife_output.index.year
 
             logger.info('Jackknife on year: {}.'.format(str(this_year)))
-            train_input = jackknife_input[
-                input_years != this_year]
-            train_output = jackknife_output[
-                output_years != this_year]
-            test_input = jackknife_input[input_years ==
-                                         this_year]
-            test_output = jackknife_output[output_years ==
-                                           this_year]
+            train_input = jackknife_input[input_years != this_year]
+            train_output = jackknife_output[output_years != this_year]
+            test_input = jackknife_input[input_years == this_year]
+            test_output = jackknife_output[output_years == this_year]
 
             # Execute training
             training = NNTrain(train_input, train_output)