def test_ets_full_seasonal_fit(synthetic_data, estimator_type):
train_df, test_df, coef = synthetic_data
ets = ETSFull(response_col='response',
date_col='week',
prediction_percentiles=[5, 95],
seasonality=52,
num_warmup=50,
num_sample=50,
verbose=False,
estimator_type=estimator_type)
ets.fit(train_df)
init_call = ets.get_init_values()
assert isinstance(init_call, ETSInitializer)
assert init_call.s == 52
init_values = init_call()
assert init_values['init_sea'].shape == (51, )
predict_df = ets.predict(test_df)
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
expected_shape = (51, len(expected_columns))
expected_num_parameters = 5
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert len(ets._posterior_samples) == expected_num_parameters
assert ets.is_fitted()
def test_ets_full_reproducibility(synthetic_data, estimator_type, seasonality):
train_df, test_df, coef = synthetic_data
ets1 = ETSFull(response_col='response',
date_col='week',
prediction_percentiles=[5, 95],
seasonality=seasonality,
num_warmup=50,
num_sample=50,
verbose=False,
estimator_type=estimator_type)
# first fit and predict
ets1.fit(train_df)
posteriors1 = copy(ets1._posterior_samples)
prediction1 = ets1.predict(test_df)
# second fit and predict
# note a new instance must be created to reset the seed
# note both fit and predict contain random generation processes
ets2 = ETSFull(response_col='response',
date_col='week',
prediction_percentiles=[5, 95],
seasonality=seasonality,
num_warmup=50,
num_sample=50,
verbose=False,
estimator_type=estimator_type)
ets2.fit(train_df)
posteriors2 = copy(ets2._posterior_samples)
prediction2 = ets2.predict(test_df)
# assert same posterior keys
assert set(posteriors1.keys()) == set(posteriors2.keys())
# assert posterior draws are reproducible
for k, v in posteriors1.items():
assert np.allclose(posteriors1[k], posteriors2[k])
# assert prediction is reproducible
assert prediction1.equals(prediction2)
def test_ets_non_seasonal_fit(synthetic_data, estimator_type):
train_df, test_df, coef = synthetic_data
ets = ETSFull(
response_col='response',
date_col='week',
estimator_type=estimator_type,
num_warmup=50,
)
ets.fit(train_df)
predict_df = ets.predict(test_df)
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
expected_shape = (51, len(expected_columns))
expected_num_parameters = 4
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert len(ets._posterior_samples) == expected_num_parameters
def test_full_prediction_percentiles(iclaims_training_data,
prediction_percentiles):
df = iclaims_training_data
ets = ETSFull(
response_col='claims',
date_col='week',
seasonality=52,
seed=8888,
prediction_percentiles=prediction_percentiles,
)
if not prediction_percentiles:
p_labels = ['_5', '', '_95']
else:
p_labels = ['_5', '_10', '', '_95']
ets.fit(df)
predicted_df = ets.predict(df)
expected_columns = ['week'] + ["prediction" + p for p in p_labels]
assert predicted_df.columns.tolist() == expected_columns
assert predicted_df.shape[0] == df.shape[0]
predicted_df = ets.predict(df, decompose=True)
plot_components = [
'prediction', PredictedComponents.TREND.value,
PredictedComponents.SEASONALITY.value
]
expected_columns = ['week']
for pc in plot_components:
for p in p_labels:
expected_columns.append(pc + p)
assert predicted_df.columns.tolist() == expected_columns
assert predicted_df.shape[0] == df.shape[0]