Skip to content

Commit

Permalink
Update test_all_forecasters.py
Browse files Browse the repository at this point in the history
  • Loading branch information
@fkiraly
fkiraly committed May 3, 2024
1 parent c04513c commit f47d0eb
Showing 1 changed file with 41 additions and 20 deletions.
61 changes: 41 additions & 20 deletions sktime/forecasting/tests/test_all_forecasters.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,8 @@
"is currently experimental and not supported everywhere"
)

RAND_SEED = 42


class ForecasterFixtureGenerator(BaseFixtureGenerator):
"""Fixture generator for forecasting tests.
Expand Down Expand Up @@ -170,7 +172,7 @@ def test_raises_not_fitted_error(self, estimator_instance):
def test_y_multivariate_raises_error(self, estimator_instance):
"""Test that wrong y scitype raises error (uni/multivariate not supported)."""
if estimator_instance.get_tag("scitype:y") == "multivariate":
y = _make_series(n_columns=1)
y = _make_series(n_columns=1, random_state=RAND_SEED)
with pytest.raises(ValueError, match=r"two or more variables"):
estimator_instance.fit(y, fh=FH0)

Expand All @@ -190,7 +192,7 @@ def test_y_invalid_type_raises_error(self, estimator_instance, y):
@pytest.mark.parametrize("X", INVALID_X_INPUT_TYPES)
def test_X_invalid_type_raises_error(self, estimator_instance, n_columns, X):
"""Test that invalid X input types raise error."""
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)
try:
with pytest.raises(TypeError, match=r"type"):
estimator_instance.fit(y_train, X, fh=FH0)
Expand Down Expand Up @@ -219,7 +221,10 @@ def test_predict_time_index(
# "is currently experimental and not supported everywhere"
# )
y_train = _make_series(
n_columns=n_columns, index_type=index_type, n_timepoints=50
n_columns=n_columns,
index_type=index_type,
n_timepoints=50,
random_state=RAND_SEED,
)
cutoff = get_cutoff(y_train, return_index=True)
fh = _make_fh(cutoff, fh_int, fh_type, is_relative)
Expand All @@ -237,7 +242,10 @@ def test_predict_time_index(
_assert_correct_columns(y_pred, y_train)

y_test = _make_series(
n_columns=n_columns, index_type=index_type, n_timepoints=len(y_pred)
n_columns=n_columns,
index_type=index_type,
n_timepoints=len(y_pred),
random_state=RAND_SEED,
)
y_test.index = y_pred.index
y_res = estimator_instance.predict_residuals(y_test)
Expand Down Expand Up @@ -278,7 +286,9 @@ def test_predict_time_index_with_X(

# Some estimators may not support all time index types and fh types, hence we
# need to catch NotImplementedErrors.
y = _make_series(n_columns=n_columns, index_type=index_type)
y = _make_series(
n_columns=n_columns, index_type=index_type, random_state=RAND_SEED
)
cutoff = get_cutoff(y.iloc[: len(y) // 2], return_index=True)
fh = _make_fh(cutoff, fh_int_oos, fh_type, is_relative)

Expand Down Expand Up @@ -314,7 +324,9 @@ def test_predict_time_index_in_sample_full(
# "ForecastingHorizon with timedelta values "
# "is currently experimental and not supported everywhere"
# )
y_train = _make_series(n_columns=n_columns, index_type=index_type)
y_train = _make_series(
n_columns=n_columns, index_type=index_type, random_state=RAND_SEED
)
cutoff = get_cutoff(y_train, return_index=True)
steps = -np.arange(len(y_train))
fh = _make_fh(cutoff, steps, fh_type, is_relative)
Expand All @@ -339,7 +351,7 @@ def test_predict_series_name_preserved(self, estimator_instance):
if estimator_instance.get_tag("scitype:y") == "multivariate":
return None

y_train = _make_series(n_timepoints=15)
y_train = _make_series(n_timepoints=15, random_state=RAND_SEED)
y_train.name = "foo"

estimator_instance.fit(y_train, fh=[1, 2, 3])
Expand Down Expand Up @@ -417,7 +429,9 @@ def test_predict_interval(
AssertionError - if Forecaster test instance does not have "capability:pred_int"
and no NotImplementedError is raised when calling predict_interval
"""
y_train = _make_series(n_columns=n_columns, index_type=index_type)
y_train = _make_series(
n_columns=n_columns, index_type=index_type, random_state=RAND_SEED
)
estimator_instance.fit(y_train, fh=fh_int_oos)
if estimator_instance.get_tag("capability:pred_int"):
pred_ints = estimator_instance.predict_interval(
Expand Down Expand Up @@ -506,7 +520,7 @@ def test_predict_quantiles(self, estimator_instance, n_columns, fh_int_oos, alph
AssertionError - if Forecaster test instance does not have "capability:pred_int"
and no NotImplementedError is raised when calling predict_quantiles
"""
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)
estimator_instance.fit(y_train, fh=fh_int_oos)
if estimator_instance.get_tag("capability:pred_int"):
quantiles = estimator_instance.predict_quantiles(fh=fh_int_oos, alpha=alpha)
Expand Down Expand Up @@ -559,7 +573,7 @@ def test_predict_proba(self, estimator_instance, n_columns, fh_int_oos):
AssertionError - if Forecaster test instance does not have "capability:pred_int"
and no NotImplementedError is raised when calling predict_proba
"""
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)
estimator_instance.fit(y_train, fh=fh_int_oos)

if estimator_instance.get_tag("capability:pred_int"):
Expand Down Expand Up @@ -632,7 +646,7 @@ def test_pred_int_tag(self, estimator_instance):
)
def test_score(self, estimator_instance, n_columns, fh_int_oos):
"""Check score method."""
y = _make_series(n_columns=n_columns)
y = _make_series(n_columns=n_columns, random_state=RAND_SEED)
y_train, y_test = temporal_train_test_split(y)
estimator_instance.fit(y_train, fh=fh_int_oos)
y_pred = estimator_instance.predict()
Expand All @@ -653,7 +667,7 @@ def test_update_predict_single(
self, estimator_instance, n_columns, fh_int_oos, update_params
):
"""Check correct time index of update-predict."""
y = _make_series(n_columns=n_columns)
y = _make_series(n_columns=n_columns, random_state=RAND_SEED)
y_train, y_test = temporal_train_test_split(y)
estimator_instance.fit(y_train, fh=fh_int_oos)
y_pred = estimator_instance.update_predict_single(
Expand All @@ -677,7 +691,12 @@ def test_update_predict_predicted_index(
update_params,
):
"""Check predicted index in update_predict."""
y = _make_series(n_columns=n_columns, all_positive=True, index_type="datetime")
y = _make_series(
n_columns=n_columns,
all_positive=True,
index_type="datetime",
random_state=RAND_SEED,
)
y_train, y_test = temporal_train_test_split(y)
cv = ExpandingWindowSplitter(
fh=fh_int_oos,
Expand All @@ -700,7 +719,7 @@ def test__y_and_cutoff(self, estimator_instance, n_columns):
# check _y and cutoff is None after construction
f = estimator_instance

y = _make_series(n_columns=n_columns)
y = _make_series(n_columns=n_columns, random_state=RAND_SEED)
y_train, y_test = temporal_train_test_split(y, train_size=0.75)

# check that _y and cutoff are empty when estimator is constructed
Expand Down Expand Up @@ -728,7 +747,7 @@ def test__y_and_cutoff(self, estimator_instance, n_columns):

def test__y_when_refitting(self, estimator_instance, n_columns):
"""Test that _y is updated when forecaster is refitted."""
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)
estimator_instance.fit(y_train, fh=FH0)
estimator_instance.fit(y_train[3:], fh=FH0)
# using np.squeeze to make the test flexible to shape differences like
Expand All @@ -738,7 +757,7 @@ def test__y_when_refitting(self, estimator_instance, n_columns):
def test_fh_attribute(self, estimator_instance, n_columns):
"""Check fh attribute and error handling if two different fh are passed."""
f = estimator_instance
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)

f.fit(y_train, fh=FH0)
np.testing.assert_array_equal(f.fh, FH0)
Expand All @@ -756,7 +775,7 @@ def test_fh_attribute(self, estimator_instance, n_columns):
def test_fh_not_passed_error_handling(self, estimator_instance, n_columns):
"""Check that not passing fh in fit/predict raises correct error."""
f = estimator_instance
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)

if f.get_tag("requires-fh-in-fit"):
# if fh required in fit, should raise error if not passed in fit
Expand All @@ -782,7 +801,7 @@ def test_different_fh_in_fit_and_predict_error_handling(
# if fh is not required in fit, can be overwritten, should not raise error
if not f.get_tag("requires-fh-in-fit"):
return None
y_train = _make_series(n_columns=n_columns)
y_train = _make_series(n_columns=n_columns, random_state=RAND_SEED)
f.fit(y_train, fh=FH0)
np.testing.assert_array_equal(f.fh, FH0)
# changing fh during predict should raise error
Expand Down Expand Up @@ -819,13 +838,15 @@ def test_hierarchical_with_exogeneous(self, estimator_instance, n_columns):
min_timepoints=22,
max_timepoints=22,
index_type="period",
random_state=RAND_SEED,
)
X = _make_hierarchical(
hierarchy_levels=(2, 4),
n_columns=2,
min_timepoints=24,
max_timepoints=24,
index_type="period",
random_state=RAND_SEED,
)
X.columns = ["foo", "bar"]
X_train = get_window(X, lag=2)
Expand Down Expand Up @@ -884,8 +905,8 @@ def test_fit_predict(self, estimator_instance, n_columns):
Does not check directly against fit and predict, as either may
be stochastic and not return the same result each time.
"""
y = _make_series(n_columns=n_columns)
X = _make_series(n_columns=3)
y = _make_series(n_columns=n_columns, random_state=RAND_SEED)
X = _make_series(n_columns=3, random_state=RAND_SEED)

fh = ForecastingHorizon([1, 2, 3])

Expand Down

0 comments on commit f47d0eb

Please sign in to comment.