Forecast Error Measures
Scoreboard has templates for many of the common measures used for error analysis.
The scoreboards can display the error values but you can also construct the error values in a favorite view as a Calculated Measure.
Creating a scoreboard from a Formula Template is possible through the right-click option for a scoreboard record. Select from the Select Property column.
M3 DMP will display the Calculation Parameters dialog box with the formula at the bottom of the display. You can copy the formula and make it into a calculated measure in a favorite view, for example, if the scoreboard is not used when opening the favorite view or using the favorite view data manager to copy the values across favorite views.
List of common forecast error measures
| Measure | Formula |
|---|---|
| Mean Forecast Error (MFE) | MFE: Avg(3,[M1] - [M2]) |
| Mean Absolute Deviation (MAD) | MAD: Avg(3,Abs([M1] - [M2])) |
| Mean Absolute Percentage Error (MAPE) | MAPE: Avg(3,Abs(([M1] - [M2])
/ [M1])) This is a common and useful indicator that gives relative information about the difference between two measures. Demand vs. Forecast or Forecast vs Sales, and so on. |
| Mean Squared Error (MSE) | MSE: Avg(3,([M1] - [M2])^2) |
| Root Mean Squared Error (RMSE) | RMSE: Sqr(Avg(3,([M1] - [M2])^2)) |
| Tracking Signal (TS) | TS: Sum(3,([M1] - [M2])) /
Abs(([M1] - [M2])) Tracking signal is used for accumulating errors over time to detect when the basic pattern has changed. Typically, the forecaster needs to be notified when the tracking signal falls outside certain limits so that you can apply a new forecast model or a manual correction. |
| Accuracy Signal (AS) | AS: ([M1] - [M2]) / ([M1] +
[M2]) Accuracy Signal gives an indicator to how close we are to a perfect forecast and if we are overshooting or undershooting.
Note: AS has the same scale for positive and
negative values.
|
| Theil's U-statistic (Theil's U) | Theil's U: Sqr(Sum(3,(([M1][P
+ 1] - [M2][P + 1]) / [M2][P])^2)/Sum(3,(([M2][P + 1] - [M2][P]) /
[M2][P])^2)) This provides a relative comparison and also squares the errors involved so that large errors are given more weight than small errors. Theil's U describes how good a fit two measures are.
|