Skip to contents

Measures the quality of quantile (probabilistic) forecasts using the pinball loss, also known as the quantile loss. The loss is averaged over all observations and all predicted quantile levels. Smaller scores indicate better calibrated quantile forecasts.

Details

For a single quantile level \(\tau\) with forecast \(q_i\) and observation \(y_i\) the pinball loss is $$ L_\tau(y_i, q_i) = \begin{cases} \tau\,(y_i - q_i), & \text{if } y_i \ge q_i \\ (1 - \tau)\,(q_i - y_i), & \text{if } y_i < q_i \end{cases} $$ The reported score is twice the mean of \(L_\tau\) over all observations and all quantile levels \(\tau\), matching the convention used by fabletools so that the median (\(\tau = 0.5\)) pinball loss equals the mean absolute error.

Dictionary

This mlr3::Measure can be instantiated via the dictionary mlr3::mlr_measures or with the associated sugar function mlr3::msr():

mlr_measures$get("fcst.pinball")
msr("fcst.pinball")

Meta Information

  • Task type: “regr”

  • Range: \([0, \infty)\)

  • Minimize: TRUE

  • Average: macro

  • Required Prediction: “quantiles”

  • Required Packages: mlr3, mlr3forecast

Parameters

Empty ParamSet

References

Koenker, Roger, Bassett, Gilbert (1978). “Regression Quantiles.” Econometrica, 46(1), 33–50.

See also

Other Measure: mlr_measures_fcst.acf1, mlr_measures_fcst.coverage, mlr_measures_fcst.mase, mlr_measures_fcst.mda, mlr_measures_fcst.mdpv, mlr_measures_fcst.mdv, mlr_measures_fcst.mpe, mlr_measures_fcst.msis, mlr_measures_fcst.rmsse, mlr_measures_fcst.wape, mlr_measures_fcst.winkler

Super classes

mlr3::Measure -> mlr3::MeasureRegr -> MeasurePinball

Methods

Inherited methods


MeasurePinball$new()

Creates a new instance of this R6 class.

Usage


MeasurePinball$clone()

The objects of this class are cloneable with this method.

Usage

MeasurePinball$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.