Add inverse transformation into predict method of pipelines (#1314)

CHANGELOG.md

@@ -26,6 +26,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - 
 - `mrmr` feature selection working with categoricals ([#1311](https://github.com/tinkoff-ai/etna/pull/1311))
 - Fix version of `statsforecast` to 1.4 to avoid dependency conflicts during installation ([#1313](https://github.com/tinkoff-ai/etna/pull/1313))
+- Add inverse transformation into `predict` method of pipelines ([#1314](https://github.com/tinkoff-ai/etna/pull/1314))
 
 ### Removed
 - Building docker images with cuda 10.2 ([#1306](https://github.com/tinkoff-ai/etna/pull/1306))

etna/pipeline/mixins.py

@@ -105,6 +105,8 @@ def _predict(
             )
         else:
             raise NotImplementedError(f"Unknown model type: {self.model.__class__.__name__}!")
+
+        results.inverse_transform(self.transforms)
         return results
 
 

tests/test_pipeline/test_autoregressive_pipeline.py

@@ -260,7 +260,7 @@ def test_backtest_forecasts_sanity(step_ts: TSDataset):
         (ProphetModel(), []),
     ],
 )
-def test_predict(model, transforms, example_tsds):
+def test_predict_format(model, transforms, example_tsds):
     ts = example_tsds
     pipeline = AutoRegressivePipeline(model=model, transforms=transforms, horizon=7)
     pipeline.fit(ts)
@@ -282,6 +282,23 @@ def test_predict(model, transforms, example_tsds):
     assert len(result_df) == len(example_tsds.segments) * num_points
 
 
+def test_predict_values(example_tsds):
+    original_ts = deepcopy(example_tsds)
+
+    model = LinearPerSegmentModel()
+    transforms = [AddConstTransform(in_column="target", value=10, inplace=True), DateFlagsTransform()]
+    pipeline = AutoRegressivePipeline(model=model, transforms=transforms, horizon=5)
+    pipeline.fit(example_tsds)
+    predictions_pipeline = pipeline.predict(ts=original_ts)
+
+    original_ts.fit_transform(transforms)
+    model.fit(original_ts)
+    predictions_manual = model.predict(original_ts)
+    predictions_manual.inverse_transform(transforms)
+
+    pd.testing.assert_frame_equal(predictions_pipeline.to_pandas(), predictions_manual.to_pandas())
+
+
 @pytest.mark.parametrize("load_ts", [True, False])
 @pytest.mark.parametrize(
     "model, transforms",

tests/test_pipeline/test_mixins.py

@@ -113,6 +113,26 @@ def test_predict_mixin_predict_create_ts_called(start_timestamp, end_timestamp,
     mixin._create_ts.assert_called_once_with(ts=ts, start_timestamp=start_timestamp, end_timestamp=end_timestamp)
 
 
+@pytest.mark.parametrize(
+    "start_timestamp, end_timestamp",
+    [
+        (pd.Timestamp("2020-01-01"), pd.Timestamp("2020-01-01")),
+        (pd.Timestamp("2020-01-01"), pd.Timestamp("2020-01-02")),
+        (pd.Timestamp("2020-01-01"), pd.Timestamp("2020-01-10")),
+        (pd.Timestamp("2020-01-05"), pd.Timestamp("2020-01-10")),
+    ],
+)
+def test_predict_mixin_predict_inverse_transform_called(start_timestamp, end_timestamp, example_tsds):
+    ts = MagicMock()
+    mixin = make_mixin()
+
+    result = mixin._predict(
+        ts=ts, start_timestamp=start_timestamp, end_timestamp=end_timestamp, prediction_interval=False, quantiles=[]
+    )
+
+    result.inverse_transform.assert_called_once_with(mixin.transforms)
+
+
 @pytest.mark.parametrize(
     "start_timestamp, end_timestamp",
     [

tests/test_pipeline/test_pipeline.py

@@ -205,8 +205,8 @@ def test_forecast_with_intervals_other_model(base_forecast, model_class):
     )
 
 
-def test_forecast(example_tsds):
-    """Test that the forecast from the Pipeline is correct."""
+def test_forecast_values(example_tsds):
+    """Test that the forecast from the Pipeline generates correct values."""
     original_ts = deepcopy(example_tsds)
 
     model = LinearPerSegmentModel()
@@ -221,7 +221,7 @@ def test_forecast(example_tsds):
     forecast_manual = model.forecast(future)
     forecast_manual.inverse_transform(transforms)
 
-    assert np.all(forecast_pipeline.df.values == forecast_manual.df.values)
+    pd.testing.assert_frame_equal(forecast_pipeline.to_pandas(), forecast_manual.to_pandas())
 
 
 @pytest.mark.parametrize(
@@ -1121,7 +1121,7 @@ def test_pipeline_with_deepmodel(example_tsds):
         (ProphetModel(), []),
     ],
 )
-def test_predict(model, transforms, example_tsds):
+def test_predict_format(model, transforms, example_tsds):
     ts = example_tsds
     pipeline = Pipeline(model=model, transforms=transforms, horizon=7)
     pipeline.fit(ts)
@@ -1143,6 +1143,23 @@ def test_predict(model, transforms, example_tsds):
     assert len(result_df) == len(example_tsds.segments) * num_points
 
 
+def test_predict_values(example_tsds):
+    original_ts = deepcopy(example_tsds)
+
+    model = LinearPerSegmentModel()
+    transforms = [AddConstTransform(in_column="target", value=10, inplace=True), DateFlagsTransform()]
+    pipeline = Pipeline(model=model, transforms=transforms, horizon=5)
+    pipeline.fit(example_tsds)
+    predictions_pipeline = pipeline.predict(ts=original_ts)
+
+    original_ts.fit_transform(transforms)
+    model.fit(original_ts)
+    predictions_manual = model.predict(original_ts)
+    predictions_manual.inverse_transform(transforms)
+
+    pd.testing.assert_frame_equal(predictions_pipeline.to_pandas(), predictions_manual.to_pandas())
+
+
 @pytest.mark.parametrize("load_ts", [True, False])
 @pytest.mark.parametrize(
     "model, transforms",