Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[BUG] fix broken DynamicFactor._predict_interval #4479

Merged
merged 6 commits into from Apr 21, 2023
Merged

[BUG] fix broken DynamicFactor._predict_interval #4479

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
b520ed7
Update dynamic_factor.py
fkiraly Apr 16, 2023
781682c
remove test exception
fkiraly Apr 16, 2023
7b554f4
linting and bugfix
fkiraly Apr 16, 2023
524cfcc
test params
fkiraly Apr 16, 2023
6ca17d1
Merge branch 'main' into bugfix-dynamicfactor
fkiraly Apr 21, 2023
832ed3c
Merge branch 'main' into bugfix-dynamicfactor
fkiraly Apr 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
22 changes: 9 additions & 13 deletions sktime/forecasting/dynamic_factor.py
View file
Open in desktop
Expand Up @@ -268,23 +268,23 @@ def _predict_interval(self, fh, X=None, coverage: [float] = None):
# statsmodels requires zero-based indexing starting at the
# beginning of the training series when passing integers
start, end = fh.to_absolute_int(self._y.index[0], self.cutoff)[[0, -1]]
steps = end - len(self._y) + 1
ix = fh.to_indexer()

model = self._fitted_forecaster

df_list = []
# generate the forecasts for each alpha/coverage
for coverage in coverage_list:

alpha = -0.5 * coverage + 0.5
alpha = 1 - coverage

if "exog" in inspect.signature(model.__init__).parameters.keys():
y_pred = model.get_forecast(steps=(end - start) + 1, exog=X).conf_int(
alpha=alpha
)
y_pred = model.get_forecast(steps=steps, exog=X).conf_int(alpha=alpha)
else:
y_pred = model.get_forecast(steps=(end - start) + 1).conf_int(
alpha=alpha
)
y_pred = model.get_forecast(steps=steps).conf_int(alpha=alpha)

y_pred = y_pred.iloc[ix]

y_pred.rename(
columns={orig_col: orig_col + f" {coverage}" for orig_col in y_pred},
Expand Down Expand Up @@ -332,11 +332,7 @@ def _predict_interval(self, fh, X=None, coverage: [float] = None):
predictions_df_2.values, columns=pd.MultiIndex.from_tuples(final_columns)
)

if "int" in (self._y.index[0]).__class__.__name__: # Rather fishy solution
predictions_df_3.index = np.arange(
start + self._y.index[0], end + self._y.index[0] + 1
)
return predictions_df_3.loc[fh.to_absolute(self.cutoff).to_pandas()]
predictions_df_3.index = fh.to_absolute(self.cutoff).to_pandas()

return predictions_df_3

Expand Down Expand Up @@ -556,6 +552,6 @@ def get_test_params(cls, parameter_set="default"):
`create_test_instance` uses the first (or only) dictionary in `params
"""
params1 = {"k_factors": 1, "factor_order": 1}
params2 = {"enforce_stationarity": False, "maxiter": 25, "low_memory": True}
params2 = {"maxiter": 25, "low_memory": True}

return [params1, params2]
6 changes: 0 additions & 6 deletions sktime/tests/_config.py
View file
Open in desktop
Expand Up @@ -130,12 +130,6 @@
# SAX returns strange output format
# this needs to be fixed, was not tested previously due to legacy exception
"SAX": "test_fit_transform_output",
# known bug in DynamicFactor, returns wrong index, #4362
"DynamicFactor": [
"test_predict_interval",
"test_predict_quantiles",
"test_predict_proba",
],
}

# We use estimator tags in addition to class hierarchies to further distinguish
Expand Down