Weighted Model Comparison Selection Criteria

Set the selection criteria for choosing the champion pipeline on the Pipeline Comparison tab and for choosing the champion model within a pipeline.

Note: For projects using the External Forecast pipeline, if the values are missing from the ACTUAL column, the weighted metrics will have missing values in the Pipeline Comparison results. See Working with External Forecast Projects for more information about the composition of external forecast projects.

The weighted version of a statistic of fita statistical value that is used to evaluate how well a forecasting model fits the historical series by comparing the actual data to the predicted values. is calculated using the input data to assign weights to each series, with higher weights being assigned to series with higher volume. For the following weighted statistics, series weight is calculated as follows:

Series weight = StartAbsoluteValue StartFraction 1 Over upper T EndFraction normal upper Sigma Subscript t equals 1 Superscript upper T Baseline y Subscript t Baseline EndAbsoluteValue. Click image for alternative formats.

For the description of the unweighted statistics of fit, see Descriptions of Model Selection Criteria.

Weighted mean absolute percent error (WMAPE)

The weighted mean of the MAPE values for each time seriesan aggregation of transactional data into specified time intervals and sorted according to unique combinations of the default attributes (BY variables), StartFraction normal upper Sigma upper M upper A upper P upper E times upper S e r i e s upper W e i g h t Over normal upper Sigma upper S e r i e s upper W e i g h t EndFraction. Click image for alternative formats.

Weighted mean absolute error (WMAE)

The weighted mean of the MAE values for each time series, StartFraction normal upper Sigma upper M upper A upper E times upper S e r i e s upper W e i g h t Over normal upper Sigma upper S e r i e s upper W e i g h t EndFraction. Click image for alternative formats.

Weighted mean absolute scaled error (WMASE)

The weighted mean of the MASE values for each time series, StartFraction normal upper Sigma upper M upper A upper S upper E times upper S e r i e s upper W e i g h t Over normal upper Sigma upper S e r i e s upper W e i g h t EndFraction. Click image for alternative formats.

Weighted root mean squared error (WRMSE)

The weighted mean of the RMSE values for each time series, StartFraction normal upper Sigma upper R upper M upper S upper E times upper S e r i e s upper W e i g h t Over normal upper Sigma upper S e r i e s upper W e i g h t EndFraction. Click image for alternative formats.

Weighted absolute percent error (WAPE)

The weighted mean of the APE values for each time series, StartFraction normal upper Sigma Subscript t StartAbsoluteValue upper E upper R upper R upper O upper R EndAbsoluteValue Subscript t Baseline Over normal upper Sigma Subscript t StartAbsoluteValue upper A upper C upper T upper U upper A upper L EndAbsoluteValue Subscript t Baseline EndFraction. Click image for alternative formats.

See Notations Used in the Formulas for a description of the APE formula.

Weighted absolute scaled error (WASE)

The weighted mean of the ASE values for each time series, StartStartFraction normal upper Sigma Subscript t StartAbsoluteValue upper A upper C upper T upper U upper A upper L EndAbsoluteValue Subscript t Baseline times times StartFraction StartAbsoluteValue upper E upper R upper R upper O upper R EndAbsoluteValue t Subscript Baseline Over StartAbsoluteValue upper M upper A upper E upper R upper R upper O upper R EndAbsoluteValue t Subscript Baseline EndFraction OverOver normal upper Sigma t Subscript StartAbsoluteValue upper A upper C upper T upper U upper A upper L EndAbsoluteValue Subscript t Baseline EndEndFraction. Click image for alternative formats.

See Also

Last updated: March 16, 2026