autogluon.timeseries 1.1.2b20241118__py3-none-any.whl → 1.1.2b20241119__py3-none-any.whl

autogluon/timeseries/metrics/point.py

@@ -8,7 +8,7 @@ import pandas as pd
 from autogluon.timeseries import TimeSeriesDataFrame
 
 from .abstract import TimeSeriesScorer
-from .utils import _in_sample_abs_seasonal_error, _in_sample_squared_seasonal_error
+from .utils import in_sample_abs_seasonal_error, in_sample_squared_seasonal_error
 
 logger = logging.getLogger(__name__)
 
@@ -232,7 +232,7 @@ class MASE(TimeSeriesScorer):
     def save_past_metrics(
         self, data_past: TimeSeriesDataFrame, target: str = "target", seasonal_period: int = 1, **kwargs
     ) -> None:
-        self._past_abs_seasonal_error = _in_sample_abs_seasonal_error(
+        self._past_abs_seasonal_error = in_sample_abs_seasonal_error(
             y_past=data_past[target], seasonal_period=seasonal_period
         )
 
@@ -292,7 +292,7 @@ class RMSSE(TimeSeriesScorer):
     def save_past_metrics(
         self, data_past: TimeSeriesDataFrame, target: str = "target", seasonal_period: int = 1, **kwargs
     ) -> None:
-        self._past_squared_seasonal_error = _in_sample_squared_seasonal_error(
+        self._past_squared_seasonal_error = in_sample_squared_seasonal_error(
             y_past=data_past[target], seasonal_period=seasonal_period
         )
 

autogluon/timeseries/metrics/quantile.py

@@ -6,7 +6,7 @@ import pandas as pd
 from autogluon.timeseries.dataset.ts_dataframe import TimeSeriesDataFrame
 
 from .abstract import TimeSeriesScorer
-from .utils import _in_sample_abs_seasonal_error
+from .utils import in_sample_abs_seasonal_error
 
 
 class WQL(TimeSeriesScorer):
@@ -85,7 +85,7 @@ class SQL(TimeSeriesScorer):
     def save_past_metrics(
         self, data_past: TimeSeriesDataFrame, target: str = "target", seasonal_period: int = 1, **kwargs
     ) -> None:
-        self._past_abs_seasonal_error = _in_sample_abs_seasonal_error(
+        self._past_abs_seasonal_error = in_sample_abs_seasonal_error(
             y_past=data_past[target], seasonal_period=seasonal_period
         )
 

autogluon/timeseries/metrics/utils.py

@@ -7,12 +7,12 @@ def _get_seasonal_diffs(*, y_past: pd.Series, seasonal_period: int = 1) -> pd.Se
     return y_past.groupby(level=ITEMID, sort=False).diff(seasonal_period).abs()
 
 
-def _in_sample_abs_seasonal_error(*, y_past: pd.Series, seasonal_period: int = 1) -> pd.Series:
+def in_sample_abs_seasonal_error(*, y_past: pd.Series, seasonal_period: int = 1) -> pd.Series:
     """Compute seasonal naive forecast error (predict value from seasonal_period steps ago) for each time series."""
     seasonal_diffs = _get_seasonal_diffs(y_past=y_past, seasonal_period=seasonal_period)
     return seasonal_diffs.groupby(level=ITEMID, sort=False).mean().fillna(1.0)
 
 
-def _in_sample_squared_seasonal_error(*, y_past: pd.Series, seasonal_period: int = 1) -> pd.Series:
+def in_sample_squared_seasonal_error(*, y_past: pd.Series, seasonal_period: int = 1) -> pd.Series:
     seasonal_diffs = _get_seasonal_diffs(y_past=y_past, seasonal_period=seasonal_period)
     return seasonal_diffs.pow(2.0).groupby(level=ITEMID, sort=False).mean().fillna(1.0)

autogluon/timeseries/models/autogluon_tabular/mlforecast.py

@@ -11,6 +11,7 @@ from sklearn.base import BaseEstimator
 import autogluon.core as ag
 from autogluon.tabular import TabularPredictor
 from autogluon.timeseries.dataset.ts_dataframe import ITEMID, TIMESTAMP, TimeSeriesDataFrame
+from autogluon.timeseries.metrics.utils import in_sample_squared_seasonal_error
 from autogluon.timeseries.models.abstract import AbstractTimeSeriesModel
 from autogluon.timeseries.models.local import SeasonalNaiveModel
 from autogluon.timeseries.utils.datetime import (
@@ -82,7 +83,6 @@ class AbstractMLForecastModel(AbstractTimeSeriesModel):
         self._mlf: Optional[MLForecast] = None
         self._scaler: Optional[BaseTargetTransform] = None
         self._residuals_std_per_item: Optional[pd.Series] = None
-        self._avg_residuals_std: Optional[float] = None
         self._train_target_median: Optional[float] = None
         self._non_boolean_real_covariates: List[str] = []
 
@@ -332,7 +332,7 @@ class AbstractMLForecastModel(AbstractTimeSeriesModel):
     def _save_residuals_std(self, val_df: pd.DataFrame) -> None:
         """Compute standard deviation of residuals for each item using the validation set.
 
-        Saves per-item residuals to `self.residuals_std_per_item` and average std to `self._avg_residuals_std`.
+        Saves per-item residuals to `self.residuals_std_per_item`.
         """
         residuals_df = val_df[[MLF_ITEMID, MLF_TARGET]]
         residuals_df = residuals_df.assign(y_pred=self._mlf.models_["mean"].predict(val_df))
@@ -344,7 +344,6 @@ class AbstractMLForecastModel(AbstractTimeSeriesModel):
         self._residuals_std_per_item = (
             residuals.pow(2.0).groupby(val_df[MLF_ITEMID].values, sort=False).mean().pow(0.5)
         )
-        self._avg_residuals_std = np.sqrt(residuals.pow(2.0).mean())
 
     def _remove_short_ts_and_generate_fallback_forecast(
         self,
@@ -391,7 +390,7 @@ class AbstractMLForecastModel(AbstractTimeSeriesModel):
             forecast_for_short_series = None
         return data_long, known_covariates_long, forecast_for_short_series
 
-    def _add_gaussian_quantiles(self, predictions: pd.DataFrame, repeated_item_ids: pd.Series):
+    def _add_gaussian_quantiles(self, predictions: pd.DataFrame, repeated_item_ids: pd.Series, past_target: pd.Series):
         """
         Add quantile levels assuming that residuals follow normal distribution
         """
@@ -403,9 +402,14 @@ class AbstractMLForecastModel(AbstractTimeSeriesModel):
         normal_scale_per_timestep = pd.Series(np.tile(sqrt_h, num_items), index=repeated_item_ids)
 
         residuals_std_per_timestep = self._residuals_std_per_item.reindex(repeated_item_ids)
-        # Use avg_residuals_std in case unseen item received for prediction
-        if residuals_std_per_timestep.isna().any():
-            residuals_std_per_timestep = residuals_std_per_timestep.fillna(value=self._avg_residuals_std)
+        # Use in-sample seasonal error in for items not seen during fit
+        items_not_seen_during_fit = residuals_std_per_timestep.index[residuals_std_per_timestep.isna()].unique()
+        if len(items_not_seen_during_fit) > 0:
+            scale_for_new_items: pd.Series = np.sqrt(
+                in_sample_squared_seasonal_error(y_past=past_target.loc[items_not_seen_during_fit])
+            )
+            residuals_std_per_timestep = residuals_std_per_timestep.fillna(scale_for_new_items)
+
         std_per_timestep = residuals_std_per_timestep * normal_scale_per_timestep
         for q in self.quantile_levels:
             predictions[str(q)] = predictions["mean"] + norm.ppf(q) * std_per_timestep.to_numpy()
@@ -493,7 +497,6 @@ class DirectTabularModel(AbstractMLForecastModel):
         if self.is_quantile_model:
             # Quantile model does not require residuals to produce prediction intervals
             self._residuals_std_per_item = pd.Series(1.0, index=val_df[MLF_ITEMID].unique())
-            self._avg_residuals_std = 1.0
         else:
             super()._save_residuals_std(val_df=val_df)
 
@@ -545,7 +548,9 @@ class DirectTabularModel(AbstractMLForecastModel):
                 predictions = apply_inverse_transform(predictions, transform=tfm)
 
         if not self.is_quantile_model:
-            predictions = self._add_gaussian_quantiles(predictions, repeated_item_ids=predictions[MLF_ITEMID])
+            predictions = self._add_gaussian_quantiles(
+                predictions, repeated_item_ids=predictions[MLF_ITEMID], past_target=data[self.target]
+            )
         predictions = TimeSeriesDataFrame(predictions.rename(columns={MLF_ITEMID: ITEMID, MLF_TIMESTAMP: TIMESTAMP}))
 
         if forecast_for_short_series is not None:
@@ -661,7 +666,9 @@ class RecursiveTabularModel(AbstractMLForecastModel):
             )
         predictions = raw_predictions.rename(columns={MLF_ITEMID: ITEMID, MLF_TIMESTAMP: TIMESTAMP})
         predictions = TimeSeriesDataFrame(
-            self._add_gaussian_quantiles(predictions, repeated_item_ids=predictions[ITEMID])
+            self._add_gaussian_quantiles(
+                predictions, repeated_item_ids=predictions[ITEMID], past_target=data[self.target]
+            )
         )
 
         if forecast_for_short_series is not None:

autogluon/timeseries/version.py

@@ -1,3 +1,3 @@
 """This is the autogluon version file."""
-__version__ = '1.1.2b20241118'
+__version__ = '1.1.2b20241119'
 __lite__ = False

{autogluon.timeseries-1.1.2b20241118.dist-info → autogluon.timeseries-1.1.2b20241119.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.1.2b20241118
+Version: 1.1.2b20241119
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community
@@ -53,9 +53,9 @@ Requires-Dist: fugue>=0.9.0
 Requires-Dist: tqdm<5,>=4.38
 Requires-Dist: orjson~=3.9
 Requires-Dist: tensorboard<3,>=2.9
-Requires-Dist: autogluon.core[raytune]==1.1.2b20241118
-Requires-Dist: autogluon.common==1.1.2b20241118
-Requires-Dist: autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241118
+Requires-Dist: autogluon.core[raytune]==1.1.2b20241119
+Requires-Dist: autogluon.common==1.1.2b20241119
+Requires-Dist: autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241119
 Provides-Extra: all
 Requires-Dist: optimum[onnxruntime]<1.20,>=1.17; extra == "all"
 Provides-Extra: chronos-onnx

{autogluon.timeseries-1.1.2b20241118.dist-info → autogluon.timeseries-1.1.2b20241119.dist-info}/RECORD

@@ -1,27 +1,27 @@
-autogluon.timeseries-1.1.2b20241118-py3.8-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
+autogluon.timeseries-1.1.2b20241119-py3.8-nspkg.pth,sha256=cQGwpuGPqg1GXscIwt-7PmME1OnSpD-7ixkikJ31WAY,554
 autogluon/timeseries/__init__.py,sha256=_CrLLc1fkjen7UzWoO0Os8WZoHOgvZbHKy46I8v_4k4,304
 autogluon/timeseries/evaluator.py,sha256=l642tYfTHsl8WVIq_vV6qhgAFVFr9UuZD7gLra3A_Kc,250
 autogluon/timeseries/learner.py,sha256=3dUxI-U6TGfNtRQUzWTvBIo1GKeXYOhxIX_q7Fed9eA,14013
 autogluon/timeseries/predictor.py,sha256=R9m-TYmlA4WoJbdYEL_AnEM26EhRIclynOfSmpO7mBk,84926
 autogluon/timeseries/regressor.py,sha256=tqQ2zWImxpuEyaAM0DeCjOZ-xcWUYZbCXsqd471xXxQ,8351
 autogluon/timeseries/splitter.py,sha256=eghGwAAN2_cxGk5aJBILgjGWtLzjxJcytMy49gg_q18,3061
-autogluon/timeseries/version.py,sha256=UHWwiMUBp4i7dutoeNAL53RPORnbkFt4q_j5O-4RsU0,90
+autogluon/timeseries/version.py,sha256=0242cKF96lnZcXGbtRmKZUi3yVY1-rAip2PtRhF89WQ,90
 autogluon/timeseries/configs/__init__.py,sha256=BTtHIPCYeGjqgOcvqb8qPD4VNX-ICKOg6wnkew1cPOE,98
 autogluon/timeseries/configs/presets_configs.py,sha256=94-yL9teDHKs2irWjP3kpewI7FE1ChYCgEgz9XHJ6gc,1965
 autogluon/timeseries/dataset/__init__.py,sha256=UvnhAN5tjgxXTHoZMQDy64YMDj4Xxa68yY7NP4vAw0o,81
 autogluon/timeseries/dataset/ts_dataframe.py,sha256=9bJQeg3HkPeVnyxzwqAJiTJGYXths7vxUV_3-OsJ6pk,48640
 autogluon/timeseries/metrics/__init__.py,sha256=LLGmYaexsx7CregV-QaHc5exjZbsJfBSVOtxHRGC0ho,2139
 autogluon/timeseries/metrics/abstract.py,sha256=9xCFQ3NaR1C0hn01M7oBd72a_CiNV-w6QFcRjwUbKYI,8183
-autogluon/timeseries/metrics/point.py,sha256=z366XJz3n4MFl4JkXOD6ZxL69F_j7Y-jbrwb7J3yDqk,15513
-autogluon/timeseries/metrics/quantile.py,sha256=owMbOAJYwVyzdRkrJpuCGUXk937GU843QndCZyp5n9Y,3967
-autogluon/timeseries/metrics/utils.py,sha256=eJ63TCR-UwbeJ1c2Qm7B2q-8B3sFthPgiooEccrf2Kc,912
+autogluon/timeseries/metrics/point.py,sha256=b19Ed4dS_ROdkrOZIik_Q3-8deCN9IQSZXtCl8tuxto,15509
+autogluon/timeseries/metrics/quantile.py,sha256=eemdLbo3y2wstnVkuA-f55YXywctUmSW1EhIW4BsoH4,3965
+autogluon/timeseries/metrics/utils.py,sha256=HuDe1BNe8yJU4f_DKM913nNrUueoRaw6zhxm1-S20s0,910
 autogluon/timeseries/models/__init__.py,sha256=MYD9JJ-wUDE5B6jW6E6LU2eXQ6vflfQBvqQJkdzJa3A,1189
 autogluon/timeseries/models/presets.py,sha256=ujNt_hft_5eNkh-Wj_Na9GBdBmI-JdnBnOEHq8X0qXc,11778
 autogluon/timeseries/models/abstract/__init__.py,sha256=wvDsQAZIV0N3AwBeMaGItoQ82trEfnT-nol2AAOIxBg,102
 autogluon/timeseries/models/abstract/abstract_timeseries_model.py,sha256=B1R0PBymUAwhIGkPIt29X-J9o9Ipdu-bQR0gK-nmcRU,30320
 autogluon/timeseries/models/abstract/model_trial.py,sha256=ENPg_7nsdxIvaNM0o0UShZ3x8jFlRmwRc5m0fGPC0TM,3720
 autogluon/timeseries/models/autogluon_tabular/__init__.py,sha256=r9i6jWcyeLHYClkcMSKRVsfrkBUMxpDrTATNTBc_qgQ,136
-autogluon/timeseries/models/autogluon_tabular/mlforecast.py,sha256=aU82AXXbpum-rtpOGGhK0r1CSYZclgDX5qG8nNl_2Mo,33018
+autogluon/timeseries/models/autogluon_tabular/mlforecast.py,sha256=vfWXLdxYlbzjKJa1qrFN-qzxgG2tiWxOSBOnmnNVxtA,33295
 autogluon/timeseries/models/autogluon_tabular/transforms.py,sha256=XVoy8KpvoeX38lHHAXq4Be9LCxKjxZ36SOFeSAICRFM,2524
 autogluon/timeseries/models/autogluon_tabular/utils.py,sha256=Fn3Vu_Q0PCtEUbtNgLp1xIblg7dOdpFlF3W5kLHgruI,63
 autogluon/timeseries/models/chronos/__init__.py,sha256=wT77HzTtmQxW3sw2k0mA5Ot6PSHivX-Uvn5fjM05EU4,60
@@ -60,11 +60,11 @@ autogluon/timeseries/utils/datetime/base.py,sha256=3NdsH3NDq4cVAOSoy3XpaNixyNlbj
 autogluon/timeseries/utils/datetime/lags.py,sha256=GoLtvcZ8oKb3QkoBJ9E59LSPLOP7Qjxrr2UmMSZgjyw,5909
 autogluon/timeseries/utils/datetime/seasonality.py,sha256=h_4w00iEytAz_N_EpCENQ8RCXy7KQITczrYjBgVqWkQ,764
 autogluon/timeseries/utils/datetime/time_features.py,sha256=PAXbYbQ0z_5GFbkxSNi41zLY_2-U3x0Ynm1m_WhdtGc,2572
-autogluon.timeseries-1.1.2b20241118.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
-autogluon.timeseries-1.1.2b20241118.dist-info/METADATA,sha256=ArxiY45x9G9GnoV9ZKmLS1yLW4M_cuO42hjwutiZ6gU,12388
-autogluon.timeseries-1.1.2b20241118.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
-autogluon.timeseries-1.1.2b20241118.dist-info/WHEEL,sha256=bFJAMchF8aTQGUgMZzHJyDDMPTO3ToJ7x23SLJa1SVo,92
-autogluon.timeseries-1.1.2b20241118.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.timeseries-1.1.2b20241118.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
-autogluon.timeseries-1.1.2b20241118.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
-autogluon.timeseries-1.1.2b20241118.dist-info/RECORD,,
+autogluon.timeseries-1.1.2b20241119.dist-info/LICENSE,sha256=CeipvOyAZxBGUsFoaFqwkx54aPnIKEtm9a5u2uXxEws,10142
+autogluon.timeseries-1.1.2b20241119.dist-info/METADATA,sha256=I9IrNzlXQskc4xcwgmngBRT6iIDZBdmd0rvXxOWaXcQ,12388
+autogluon.timeseries-1.1.2b20241119.dist-info/NOTICE,sha256=7nPQuj8Kp-uXsU0S5so3-2dNU5EctS5hDXvvzzehd7E,114
+autogluon.timeseries-1.1.2b20241119.dist-info/WHEEL,sha256=bFJAMchF8aTQGUgMZzHJyDDMPTO3ToJ7x23SLJa1SVo,92
+autogluon.timeseries-1.1.2b20241119.dist-info/namespace_packages.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.timeseries-1.1.2b20241119.dist-info/top_level.txt,sha256=giERA4R78OkJf2ijn5slgjURlhRPzfLr7waIcGkzYAo,10
+autogluon.timeseries-1.1.2b20241119.dist-info/zip-safe,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+autogluon.timeseries-1.1.2b20241119.dist-info/RECORD,,