Remove freq argument in SeasonalNaivePredictor

src/gluonts/model/seasonal_naive/_predictor.py

@@ -25,7 +25,6 @@
  LastValueImputation,
  MissingValueImputation,
 )
-from gluonts.time_feature import get_seasonality
 
 
 class SeasonalNaivePredictor(RepresentablePredictor):
@@ -43,40 +42,31 @@ class SeasonalNaivePredictor(RepresentablePredictor):
 
  Parameters
  ----------
- freq
- Frequency of the input data
  prediction_length
- Number of time points to predict
+ Number of time points to predict.
  season_length
- Length of the seasonality pattern of the input data
+ Seasonality used to make predictions.
  imputation_method
  The imputation method to use in case of missing values.
- Defaults to `LastValueImputation` which replaces each missing
+ Defaults to :py:class:`LastValueImputation` which replaces each missing
  value with the last value that was not missing.
  """
 
  @validated()
  def __init__(
  self,
- freq: str,
  prediction_length: int,
- season_length: Optional[int] = None,
+ season_length: int,
  imputation_method: Optional[
  MissingValueImputation
  ] = LastValueImputation(),
  ) -> None:
  super().__init__(prediction_length=prediction_length)
 
- assert (
- season_length is None or season_length > 0
- ), "The value of `season_length` should be > 0"
+ assert season_length > 0, "The value of `season_length` should be > 0"
 
  self.prediction_length = prediction_length
- self.season_length = (
- season_length
- if season_length is not None
- else get_seasonality(freq)
- )
+ self.season_length = season_length
  self.imputation_method = imputation_method
 
  def predict_item(self, item: DataEntry) -> Forecast:

test/ev/test_metrics_compared_to_previous_approach.py

@@ -269,7 +269,8 @@ def test_against_former_evaluator():
  )
 
  predictor = SeasonalNaivePredictor(
- prediction_length=prediction_length, freq=freq
+ prediction_length=prediction_length,
+ season_length=get_seasonality(freq),
  )
 
  quantile_levels = (0.1, 0.5, 0.9)

test/ext/naive_2/test_predictors.py

@@ -27,6 +27,7 @@
 from gluonts.ext.naive_2 import Naive2Predictor
 from gluonts.model.predictor import Predictor
 from gluonts.model.seasonal_naive import SeasonalNaivePredictor
+from gluonts.time_feature import get_seasonality
 
 
 def generate_random_dataset(
@@ -49,17 +50,24 @@ def generate_random_dataset(
 
 
 @pytest.mark.parametrize(
- "predictor_cls", [SeasonalNaivePredictor, Naive2Predictor]
+ "make_predictor",
+ [
+ lambda freq: SeasonalNaivePredictor(
+ prediction_length=PREDICTION_LENGTH,
+ season_length=SEASON_LENGTH,
+ ),
+ lambda freq: Naive2Predictor(
+ freq=freq,
+ prediction_length=PREDICTION_LENGTH,
+ season_length=SEASON_LENGTH,
+ ),
+ ],
 )
 @pytest.mark.parametrize(
  "freq", ["1min", "15min", "30min", "1H", "2H", "12H", "7D", "1W", "1M"]
 )
-def test_predictor(predictor_cls, freq: str):
- predictor = predictor_cls(
- freq=freq,
- prediction_length=PREDICTION_LENGTH,
- season_length=SEASON_LENGTH,
- )
+def test_predictor(make_predictor, freq: str):
+ predictor = make_predictor(freq)
  dataset = list(
  generate_random_dataset(
  num_ts=NUM_TS,
@@ -87,7 +95,7 @@ def test_predictor(predictor_cls, freq: str):
  assert forecast.start_date == forecast_start(data)
 
  # specifically for the seasonal naive we can test the supposed result directly
- if predictor_cls == SeasonalNaivePredictor:
+ if isinstance(predictor, SeasonalNaivePredictor):
  assert np.allclose(forecast.samples[0], ref)
 
 
@@ -115,11 +123,25 @@ def naive_2_predictor():
 
 @flaky(max_runs=3, min_passes=1)
 @pytest.mark.parametrize(
- "predictor_cls, parameters, accuracy",
- [seasonal_naive_predictor() + (0.0,), naive_2_predictor() + (0.0,)],
+ "predictor, accuracy",
+ [
+ (
+ SeasonalNaivePredictor(
+ prediction_length=CONSTANT_DATASET_PREDICTION_LENGTH,
+ season_length=get_seasonality(CONSTANT_DATASET_FREQ),
+ ),
+ 0.0,
+ ),
+ (
+ Naive2Predictor(
+ freq=CONSTANT_DATASET_FREQ,
+ prediction_length=CONSTANT_DATASET_PREDICTION_LENGTH,
+ ),
+ 0.0,
+ ),
+ ],
 )
-def test_accuracy(predictor_cls, parameters, accuracy):
- predictor = predictor_cls(freq=CONSTANT_DATASET_FREQ, **parameters)
+def test_accuracy(predictor, accuracy):
  agg_metrics, item_metrics = backtest_metrics(
  test_dataset=constant_test_ds,
  predictor=predictor,
@@ -133,11 +155,19 @@ def test_accuracy(predictor_cls, parameters, accuracy):
 
 
 @pytest.mark.parametrize(
- "predictor_cls, parameters",
- [seasonal_naive_predictor(), naive_2_predictor()],
+ "predictor",
+ [
+ SeasonalNaivePredictor(
+ prediction_length=CONSTANT_DATASET_PREDICTION_LENGTH,
+ season_length=get_seasonality(CONSTANT_DATASET_FREQ),
+ ),
+ Naive2Predictor(
+ freq=CONSTANT_DATASET_FREQ,
+ prediction_length=CONSTANT_DATASET_PREDICTION_LENGTH,
+ ),
+ ],
 )
-def test_seriali_predictors(predictor_cls, parameters):
- predictor = predictor_cls(freq=CONSTANT_DATASET_FREQ, **parameters)
+def test_seriali_predictors(predictor):
  with tempfile.TemporaryDirectory() as temp_dir:
  predictor.serialize(Path(temp_dir))
  predictor_exp = Predictor.deserialize(Path(temp_dir))

test/model/seasonal_naive/test_seasonal_naive.py

@@ -29,7 +29,6 @@ def get_prediction(
  imputation_method=LastValueImputation(),
 ):
  pred = SeasonalNaivePredictor(
- freq=FREQ,
  prediction_length=prediction_length,
  season_length=season_length,
  imputation_method=imputation_method,
@@ -68,12 +67,9 @@ def get_prediction(
  ([1, 2, 3], [1], 1, 3, LastValueImputation()),
  ([1, 2, 3], [1, 2], 2, 3, LastValueImputation()),
  ([1, 2, 3], [1, 2, 3], 3, 3, LastValueImputation()),
- ([1, 1, 1], [1], 1, None, LastValueImputation()),
- ([1, 1, 1], [1, 1], 2, None, LastValueImputation()),
- ([1, 1, 1], [1, 1, 1], 3, None, LastValueImputation()),
- ([1, 3, np.nan], [3], 1, None, LastValueImputation()),
- ([1, 3, np.nan], [3, 3], 2, None, LastValueImputation()),
- ([1, 3, np.nan], [3, 3, 3], 3, None, LastValueImputation()),
+ ([1, 3, np.nan], [3], 1, 1, LastValueImputation()),
+ ([1, 3, np.nan], [3, 3], 2, 1, LastValueImputation()),
+ ([1, 3, np.nan], [3, 3, 3], 3, 1, LastValueImputation()),
  ([1, 3, np.nan], [np.nan], 1, 1, LeavesMissingValues()),
  ([1, 3, np.nan], [np.nan] * 2, 2, 1, LeavesMissingValues()),
  ([1, 3, np.nan], [np.nan] * 3, 3, 1, LeavesMissingValues()),

test/model/test_evaluation.py

@@ -210,9 +210,7 @@ def test_evaluate_model_vs_forecasts():
  prediction_length=3, windows=4
  )
 
- model = SeasonalNaivePredictor(
- freq="D", prediction_length=3, season_length=1
- )
+ model = SeasonalNaivePredictor(prediction_length=3, season_length=1)
 
  forecasts = list(model.predict(test_data.input))
 
@@ -246,9 +244,7 @@ def test_data_nan():
  prediction_length=3, windows=4
  )
 
- model = SeasonalNaivePredictor(
- freq="D", prediction_length=3, season_length=1
- )
+ model = SeasonalNaivePredictor(prediction_length=3, season_length=1)
 
  forecasts = list(model.predict(test_data.input))