Naive algorithm in forecast engine

When the engine is called with Naive algorithm, the initial model is fitted if the forecast for next month equals the observation for the current month, and the forecast for the subsequent month is based on the last observation repeated through the forecast horizon.

The three types of Naive model:

  • Constant Naive: The forecast considers level only. Therefore, a CONSTANT model form is used.
  • Naive with drift: The forecast considers level plus growth. Therefore, a LINEAR model form is used.
    Note: The Drift is the difference between the last and first observation divided by Number of observations.
  • Naive with seasonality: The forecast considers the seasonal profile using the most recent observations as the forecast. This uses SEASONAL or CONSTANT_SEASONAL model form.
The history and mask data of each item or location must be retrieved. See Calling forecast engines overview.
Note: For Naive with seasonality model, masks are not allowed. In case the mask data is available, the history is modelled using only the algorithm. The mask data is not considered, and a warning message is generated.

Based on the Forecast Engine, measure mapping settings retrieve the required results from the forecast engine result set and apply the results to Scenario values.

For Naive algorithm, these model forms are supported:

  • CONSTANT
  • LINEAR
  • CONSTANT_SEASONAL
  • SEASONAL

If other model forms are used, this message is displayed:

Only model forms CONSTANT, LINEAR, CONSTANT_SEASONAL and SEASONAL are supported for a call to the forecast engine using the NAIVE algorithm.

For Naive with seasonality:

  • The history data must be equal to or greater than 2 * periodicity. Else, the CONSTANT model form is used.
  • Trend/growth/drift is not calculated.