Add params_to_tune for SARIMAXModel model

CHANGELOG.md

@@ -45,6 +45,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Remove version python-3.7 from `pyproject.toml`, update lock ([#1183](https://github.com/tinkoff-ai/etna/pull/1183))
 - Add default `params_to_tune` for catboost models ([#1185](https://github.com/tinkoff-ai/etna/pull/1185))
 - Add default `params_to_tune` for `ProphetModel` ([#1203](https://github.com/tinkoff-ai/etna/pull/1203))
+- Add default `params_to_tune` for `SARIMAXModel`, change default parameters for the model ([#1206](https://github.com/tinkoff-ai/etna/pull/1206))
 ### Fixed
 - Fix bug in `GaleShapleyFeatureSelectionTransform` with wrong number of remaining features ([#1110](https://github.com/tinkoff-ai/etna/pull/1110))
 - `ProphetModel` fails with additional seasonality set ([#1157](https://github.com/tinkoff-ai/etna/pull/1157))

etna/models/sarimax.py

@@ -1,6 +1,7 @@
 import warnings
 from abc import abstractmethod
 from datetime import datetime
+from typing import Dict
 from typing import List
 from typing import Optional
 from typing import Sequence
@@ -13,13 +14,19 @@
 from statsmodels.tsa.statespace.sarimax import SARIMAXResultsWrapper
 from statsmodels.tsa.statespace.simulation_smoother import SimulationSmoother
 
+from etna import SETTINGS
 from etna.libs.pmdarima_utils import seasonal_prediction_with_confidence
 from etna.models.base import BaseAdapter
 from etna.models.base import PredictionIntervalContextIgnorantAbstractModel
 from etna.models.mixins import PerSegmentModelMixin
 from etna.models.mixins import PredictionIntervalContextIgnorantModelMixin
 from etna.models.utils import determine_num_steps
 
+if SETTINGS.auto_required:
+    from optuna.distributions import BaseDistribution
+    from optuna.distributions import CategoricalDistribution
+    from optuna.distributions import IntUniformDistribution
+
 warnings.filterwarnings(
     message="No frequency information was provided, so inferred frequency .* will be used",
     action="ignore",
@@ -374,9 +381,9 @@ class _SARIMAXAdapter(_SARIMAXBaseAdapter):
 
     def __init__(
         self,
-        order: Tuple[int, int, int] = (2, 1, 0),
-        seasonal_order: Tuple[int, int, int, int] = (1, 1, 0, 12),
-        trend: Optional[str] = "c",
+        order: Tuple[int, int, int] = (1, 0, 0),
+        seasonal_order: Tuple[int, int, int, int] = (0, 0, 0, 0),
+        trend: Optional[str] = None,
         measurement_error: bool = False,
         time_varying_regression: bool = False,
         mle_regression: bool = True,
@@ -552,9 +559,9 @@ class SARIMAXModel(
 
     def __init__(
         self,
-        order: Tuple[int, int, int] = (2, 1, 0),
-        seasonal_order: Tuple[int, int, int, int] = (1, 1, 0, 12),
-        trend: Optional[str] = "c",
+        order: Tuple[int, int, int] = (1, 0, 0),
+        seasonal_order: Tuple[int, int, int, int] = (0, 0, 0, 0),
+        trend: Optional[str] = None,
         measurement_error: bool = False,
         time_varying_regression: bool = False,
         mle_regression: bool = True,
@@ -698,3 +705,33 @@ def __init__(
                 **self.kwargs,
             )
         )
+
+    def params_to_tune(self) -> Dict[str, "BaseDistribution"]:
+        """Get default grid for tuning hyperparameters.
+
+        This grid doesn't tune ``seasonal_order.s`` parameter that determines number of periods in a season.
+        This parameter is expected to be set by the user.
+
+        Returns
+        -------
+        :
+            Grid to tune.
+        """
+        num_periods = self.seasonal_order[3]
+        if num_periods == 0:
+            return {
+                "order.0": IntUniformDistribution(low=1, high=6, step=1),
+                "order.1": IntUniformDistribution(low=1, high=2, step=1),
+                "order.2": IntUniformDistribution(low=1, high=6, step=1),
+                "trend": CategoricalDistribution(["n", "c", "t", "ct"]),
+            }
+        else:
+            return {
+                "order.0": IntUniformDistribution(low=1, high=num_periods - 1, step=1),
+                "order.1": IntUniformDistribution(low=1, high=2, step=1),
+                "order.2": IntUniformDistribution(low=1, high=num_periods - 1, step=1),
+                "seasonal_order.0": IntUniformDistribution(low=0, high=2, step=1),
+                "seasonal_order.1": IntUniformDistribution(low=0, high=1, step=1),
+                "seasonal_order.2": IntUniformDistribution(low=0, high=1, step=1),
+                "trend": CategoricalDistribution(["n", "c", "t", "ct"]),
+            }

tests/test_analysis/test_outliers/test_confidence_interval_outliers.py

@@ -51,7 +51,12 @@ def test_get_anomalies_prediction_interval_interface(outliers_tsds, model, in_co
             0.95,
             {"1": [np.datetime64("2021-01-11")], "2": [np.datetime64("2021-01-09"), np.datetime64("2021-01-27")]},
         ),
-        (SARIMAXModel, 0.999, {"1": [], "2": [np.datetime64("2021-01-27")]}),
+        (
+            SARIMAXModel,
+            {},
+            0.999,
+            {"1": [np.datetime64("2021-01-11")], "2": [np.datetime64("2021-01-09"), np.datetime64("2021-01-27")]},
+        ),
     ),
 )
 def test_get_anomalies_prediction_interval_values(outliers_tsds, model, interval_width, true_anomalies, in_column):

tests/test_models/test_sarimax_model.py

@@ -2,6 +2,7 @@
 
 import numpy as np
 import pytest
+from optuna.samplers import RandomSampler
 from statsmodels.tsa.statespace.sarimax import SARIMAXResultsWrapper
 
 from etna.models import SARIMAXModel
@@ -169,7 +170,7 @@ def test_decomposition_hamiltonian_repr_error(dfs_w_exog, components_method_name
 )
 @pytest.mark.parametrize("trend", (None, "t"))
 def test_components_names(dfs_w_exog, regressors, regressors_components, trend, components_method_name, in_sample):
-    expected_components = regressors_components + ["target_component_sarima"]
+    expected_components = regressors_components + ["target_component_arima"]
 
     train, test = dfs_w_exog
     pred_df = train if in_sample else test
@@ -236,3 +237,17 @@ def test_components_sum_up_to_target(
     components = components_method(df=pred_df)
 
     np.testing.assert_allclose(np.sum(components.values, axis=1), np.squeeze(pred))
+
+
+@pytest.mark.parametrize(
+    "model", [SARIMAXModel(seasonal_order=(0, 0, 0, 0)), SARIMAXModel(seasonal_order=(0, 0, 0, 7))]
+)
+def test_params_to_tune(model):
+    grid = model.params_to_tune()
+    # we need sampler to get a value from distribution
+    sampler = RandomSampler()
+
+    assert len(grid) > 0
+    for name, distribution in grid.items():
+        value = sampler.sample_independent(study=None, trial=None, param_name=name, param_distribution=distribution)
+        _ = model.set_params(**{name: value})