Skip to content

fix: predict with a regressor_predictor fails when training rows have a missing y #2763

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

Open
sumedhekaru wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: predict with a regressor_predictor fails when training rows have a missing y #2763

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions python/prophet/forecaster.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -818,6 +818,14 @@ def _prepare_regressors_for_predict(
df.loc[missing_hist, name] = hist_lookup.reindex(
df.loc[missing_hist, 'ds']
).to_numpy()
# Training rows with a NaN y are not in self.history, but the
# regressor model was fit on every date where the regressor is known.
missing_hist = history_mask & df[name].isna()
if missing_hist.any() and predictor.history is not None:
known = predictor.history.set_index('ds')['y']
df.loc[missing_hist, name] = known.reindex(
df.loc[missing_hist, 'ds']
).to_numpy()
if future_mask.any():
future_dates = df.loc[future_mask, 'ds']
reg_future = pd.DataFrame({'ds': future_dates})
Expand Down
16 changes: 16 additions & 0 deletions python/prophet/tests/test_prophet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1118,6 +1118,22 @@ def spy(self, df=None, vectorized=True):
assert fcst.shape[0] == 5
assert any(call == "extra" for call in calls)

def test_regressor_predictor_with_missing_y_in_history(self, backend):
df = self._make_regressor_df()[["ds", "y", "reg"]]
df.loc[[10, 11, 40], "y"] = np.nan # the regressor is known on these dates
m = Prophet(uncertainty_samples=50, stan_backend=backend)
m.add_regressor("reg", regressor_predictor=True)
m.fit(df)
future = m.make_future_dataframe(periods=5) # no regressor column
fcst = m.predict(future)
assert fcst["yhat"].notnull().all()
# History rows use the known regressor values, including the rows with a NaN y.
prepared, _ = m._prepare_regressors_for_predict(future.copy())
known = df.set_index("ds")["reg"]
np.testing.assert_allclose(
prepared["reg"].iloc[:len(df)].to_numpy(), known.to_numpy(), rtol=1e-12
)


class TestProphetWarmStart:
def test_fit_warm_start(self, daily_univariate_ts, backend):
Expand Down