REVISEDBESTFIT_FORECAST
REVISEDBESTFIT_FORECAST is an enhanced forecasting technique that selects and applies the most suitable forecasting model to improve forecast accuracy.
Logic Used for Revised Best Fit
-
Retrieve the last 12 months of sales history for the selected product and region.
Note: The 12-month period is currently hardcoded.
-
Determine the minimum and maximum sales values from the 12-month history.
-
Calculate the acceptable forecast range using the Percentage configured in the Data Measure Parameters.
Formula:
Lower Bound = Minimum Sales − (Minimum Sales × Percentage)
Upper Bound = Maximum Sales + (Maximum Sales × Percentage)
Example:
- Minimum Sales = 80
- Maximum Sales = 200
- Percentage = 20%
Lower Bound = 80 − (80 × 20%) = 64
Upper Bound = 200 + (200 × 20%) = 240
-
Sort the available Data Measures in ascending order of RMSE (based on the Error Parameter configured for Best Fit).
-
Starting with the data measure that has the lowest RMSE, verify whether the forecast value for the first forecast period falls within the calculated range.
-
If the forecast value lies within the acceptable range, that data measure is selected as the Revised Best Fit.
-
If the forecast value is outside the acceptable range, evaluate the next data measure in RMSE order and repeat the validation until a suitable data measure is found.
Selection Criteria
A data measure is selected as Revised Best Fit only when:
- It has the lowest available RMSE among the remaining candidates.
- Its first forecasted value falls within the calculated lower and upper bounds.