Prediction Interval

Two new statistical outputs from the forecast engine are added:

  • Lower Prediction Intervals
  • Upper Prediction Intervals

Prediction intervals (PIs) are upper and lower bounds around the forecast. PI represents the range of values that the actual value is considered with a specified probability. This value can be present within these bounds defined by the upper and lower prediction intervals. This probability (confidence limit) is currently set to 95%.

Some algorithms have analytical calculations to calculate the PIs while others do not. The analytical equations assume that the residuals are spread based on a normal distribution. Algorithms that do not have analytical calculations use an idea known as ‘bootstrapping’ which does not assume that the residuals are spread based on the normal or, in fact, any probability distribution.

ARIMA, BATS and Naïve algorithms use analytical calculations. All other algorithms use Bootstrapped PIs. See the documentation for more details on the different Prediction Interval calculations.

PI provides more advanced statistical output useful for additional analysis of the statistical forecast model, however this level of detail may not be applicable to all users. PIs are included in the different Forecast Graph worksheets for Demand and Inventory Planning, however, not displayed by default. PIs can be enabled by selecting the measure in the Graph’s series legend.

In Demand Planning the prediction intervals are mapped to these engines:

  • Full Monitoring Best Combination Cmonths
  • Full Monitoring Best Combination Months
  • Full Monitoring Best Combination Weeks
  • Full Monitoring Best Fit Cmonths
  • Full Monitoring Best Fit Months
  • Full Monitoring Best Fit Weeks

These engines must be used to view the prediction intervals. If prediction intervals are to be calculated by default, consider these scenarios:

  • One of these forecast engine definitions must be set to the default engine for demand via: dpls_default_forecast_engine.
  • The current default forecast engine must map these outputs to a measure.
Note: If another forecast engine definition is used from the standard content, a copy must be made before making changes.

In Inventory Planning content the prediction intervals are mapped to these engines:

  • IPLS_Bats Forecast Engine
  • IPLS_Best Fit Months
  • IPLS_Best Fit Weeks

BATS is the current default engine for Inventory.

Note: 
  • This feature is available in the forecast engine on upgrade. The related content is available after loading the demand (dpls.zip) template for this release.
  • You are not required a new role or privilege access to use this feature. It is accessible to the users with roles that have access to the aforementioned forecast engines and worksheets.