autogluon.timeseries 1.1.2b20241112__tar.gz → 1.1.2b20241113__tar.gz

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.1.2b20241112
+Version: 1.1.2b20241113
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/src/autogluon/timeseries/metrics/__init__.py

@@ -2,7 +2,7 @@ from pprint import pformat
 from typing import Type, Union
 
 from .abstract import TimeSeriesScorer
-from .point import MAE, MAPE, MASE, MSE, RMSE, RMSLE, RMSSE, SMAPE, WAPE
+from .point import MAE, MAPE, MASE, MSE, RMSE, RMSLE, RMSSE, SMAPE, WAPE, WCD
 from .quantile import SQL, WQL
 
 __all__ = [
@@ -16,6 +16,7 @@ __all__ = [
     "RMSSE",
     "SQL",
     "WAPE",
+    "WCD",
     "WQL",
 ]
 
@@ -40,6 +41,11 @@ DEPRECATED_METRICS = {
     "mean_wQuantileLoss": "WQL",
 }
 
+# Experimental metrics that are not yet user facing
+EXPERIMENTAL_METRICS = {
+    "WCD": WCD,
+}
+
 
 def check_get_evaluation_metric(
     eval_metric: Union[str, TimeSeriesScorer, Type[TimeSeriesScorer], None] = None
@@ -51,12 +57,16 @@ def check_get_evaluation_metric(
         eval_metric = eval_metric()
     elif isinstance(eval_metric, str):
         eval_metric = DEPRECATED_METRICS.get(eval_metric, eval_metric)
-        if eval_metric.upper() not in AVAILABLE_METRICS:
+        metric_name = eval_metric.upper()
+        if metric_name in AVAILABLE_METRICS:
+            eval_metric = AVAILABLE_METRICS[metric_name]()
+        elif metric_name in EXPERIMENTAL_METRICS:
+            eval_metric = EXPERIMENTAL_METRICS[metric_name]()
+        else:
             raise ValueError(
                 f"Time series metric {eval_metric} not supported. Available metrics are:\n"
                 f"{pformat(sorted(AVAILABLE_METRICS.keys()))}"
             )
-        eval_metric = AVAILABLE_METRICS[eval_metric.upper()]()
     elif eval_metric is None:
         eval_metric = AVAILABLE_METRICS[DEFAULT_METRIC_NAME]()
     else:

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/src/autogluon/timeseries/metrics/point.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 from typing import Optional
 
 import numpy as np
@@ -359,3 +360,52 @@ class RMSLE(TimeSeriesScorer):
             seasonal_period=seasonal_period,
             **kwargs,
         )
+
+
+class WCD(TimeSeriesScorer):
+    r"""Weighted cumulative discrepancy.
+
+    Measures the discrepancy between the cumulative sum of the forecast and the cumulative sum of the actual values.
+
+    .. math::
+
+        \operatorname{WCD} = 2 \cdot \frac{1}{N} \frac{1}{H} \sum_{i=1}^{N} \sum_{t=T+1}^{T+H} \alpha \cdot \max(0, -d_{i, t}) + (1 - \alpha) \cdot \max(0, d_{i, t})
+
+    where :math:`d_{i, t}` is the difference between the cumulative predicted value and the cumulative actual value
+
+    .. math::
+
+        d_{i, t} = \left(\sum_{s=T+1}^t f_{i, s}) - \left(\sum_{s=T+1}^t y_{i, s})
+
+    Parameters
+    ----------
+    alpha : float, default = 0.5
+        Values > 0.5 correspond put a stronger penalty on underpredictions (when cumulative forecast is below the
+        cumulative actual value). Values < 0.5 put a stronger penalty on overpredictions.
+    """
+
+    def __init__(self, alpha: float = 0.5):
+        assert 0 < alpha < 1, "alpha must be in (0, 1)"
+        self.alpha = alpha
+        self.num_items: Optional[int] = None
+        warnings.warn(
+            f"{self.name} is an experimental metric. Its behavior may change in the future version of AutoGluon."
+        )
+
+    def save_past_metrics(self, data_past: TimeSeriesDataFrame, **kwargs) -> None:
+        self.num_items = data_past.num_items
+
+    def _fast_cumsum(self, y: np.ndarray) -> np.ndarray:
+        """Compute the cumulative sum for each consecutive `prediction_length` items in the array."""
+        y = y.reshape(self.num_items, -1)
+        return np.nancumsum(y, axis=1).ravel()
+
+    def compute_metric(
+        self, data_future: TimeSeriesDataFrame, predictions: TimeSeriesDataFrame, target: str = "target", **kwargs
+    ) -> float:
+        y_true, y_pred = self._get_point_forecast_score_inputs(data_future, predictions, target=target)
+        cumsum_true = self._fast_cumsum(y_true.to_numpy())
+        cumsum_pred = self._fast_cumsum(y_pred.to_numpy())
+        diffs = cumsum_pred - cumsum_true
+        error = diffs * np.where(diffs < 0, -self.alpha, (1 - self.alpha))
+        return 2 * self._safemean(error)

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/src/autogluon/timeseries/version.py

@@ -1,3 +1,3 @@
 """This is the autogluon version file."""
-__version__ = '1.1.2b20241112'
+__version__ = '1.1.2b20241113'
 __lite__ = False

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/src/autogluon.timeseries.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: autogluon.timeseries
-Version: 1.1.2b20241112
+Version: 1.1.2b20241113
 Summary: Fast and Accurate ML in 3 Lines of Code
 Home-page: https://github.com/autogluon/autogluon
 Author: AutoGluon Community

{autogluon.timeseries-1.1.2b20241112 → autogluon.timeseries-1.1.2b20241113}/src/autogluon.timeseries.egg-info/requires.txt

@@ -17,9 +17,9 @@ fugue>=0.9.0
 tqdm<5,>=4.38
 orjson~=3.9
 tensorboard<3,>=2.9
-autogluon.core[raytune]==1.1.2b20241112
-autogluon.common==1.1.2b20241112
-autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241112
+autogluon.core[raytune]==1.1.2b20241113
+autogluon.common==1.1.2b20241113
+autogluon.tabular[catboost,lightgbm,xgboost]==1.1.2b20241113
 
 [all]
 optimum[onnxruntime]<1.20,>=1.17