BookmarkSubscribeRSS Feed
FVidal
Fluorite | Level 6

I would like to know if there is the possibility to change, in Forecast Studio, the MAPE indicator for a custom value. For example, It will be wonderful if we could get the DPA (Demand Planner Accuracy).

 

Thanks in advanced,

Ferran

3 REPLIES 3
alexchien
Pyrite | Level 9

HI Ferran,

As far as I know, there is not supported approach to add a custom forecast accuracy measure in Forecast Studio. You can change the measures (supported by Forecast Studio) to use in the Table view by selecting View->Select Table View Columns from the menu bar. The table sources are the OUTSTAT and RECSTAT dataset for each level of the hierarchy. If you have access to those tables, you could potentially write a post process script to modify the MAPE column (or any existing column) with whatever measure you like such as DPA (weighted APE?), and the values will be displayed in the Table view. 

thanks

Alex

FVidal
Fluorite | Level 6
Hello Alex,
Thankyou for your reply. This script that you mention can be writted an executed inside FS and see the result too in FS or I should do it in Enterprise Guide?
And, another question. Do you know if I can find, in some dataset, all the forecast information of all the models that FS ajusted not just the selected one? There is a way to get this values? It can be useful because I want to know wich of the adjusted models has the best DPA in the last 6 months. If I'm correct FS calculates MAPE for all the time serie not the lasts months.
Thanks in advanced,
Ferran
alexchien
Pyrite | Level 9

Hi Ferran,

For the custom script i mentioned, you will have to run the script outside of FS as it is not a supported process within FS. EG will do the magic.

 

 

Regarding to your second question: For each level of the hierarchy, FS produces a table called OUTSTATSELECT. This table contains the stats (the same set of stats from OUTSTAT table) for all the models considered by FS. the _SELECTED_ column shows which model is selected and the _MODEL_ column shows which model is considered for a particular row. To figure out the exact model being applied, you have to look at the SAS Catalog Autolevmodrep (in the same directory as all the other forecast outputs for each level), and take a look at the model in XML format. For example, you might see HPF0_2 as the value of the _MODEL_ column for a particular row in the OUTSTATSELECT table. You can open the SAS Catalog Autolevmodrep and copen the HPF0_2 entry. You might see something like the following:

<ARIMAXSpec version="1.0" name="HPF0_12" label="ARIMA: sale ~ D = (12) Q = (12) NOINT + INPUT1: Dif(12) EVENT2 + INPUT2: Dif(12) price" systemLabel="ARIMA: sale ~ D = (12) Q = (12) NOINT + INPUT1: Dif(12) EVENT2 + INPUT2: Dif(12) price"
<ExtendedDescription type="GENERALARIMA" source="HPFDIAGNOSE"/
<ARIMAXTable constValue="0" parmSet="0" noIntercept="1"
<EstimationOptions method="CLS" noest="0" converge="0.001" delta="0.001" gridval="0.005" maxiter="50" nols="0" nostable="0" singular="1E-7"/
<DependentInfo symbol="sale"/
<DelayDifference delay="0" numFactors="1"
<DifferenceOrder index="0" value="12"/
</DelayDifference
<Filter numPolynomials="1" type="MA"
<Polynomial index="0" type="SIMPLE" numCoefficients="1"
<Coefficient index="0" degree="12" coefficient="."/
</Polynomial
</Filter
<TransferFunction number="2"
<TFSpec index="0" symbol="EVENT2" type="EVENT"
<DelayDifference delay="0" numFactors="1"
<DifferenceOrder index="0" value="12"/
</DelayDifference
</TFSpec
<TFSpec index="1" symbol="price"
<DelayDifference delay="0" numFactors="1"
<DifferenceOrder index="0" value="12"/
</DelayDifference
</TFSpec
</TransferFunction
</ARIMAXTable
</ARIMAXSpec

This is probably as far as you can get for all the models considered in FS. Unfortunately, FS only produces forecasts for the selected model (to save storage), thus it is not possible to compute custom measures such as DPA for the unselected models.

 

Hope this helps

Alex

SAS Innovate 2025: Register Now

Registration is now open for SAS Innovate 2025 , our biggest and most exciting global event of the year! Join us in Orlando, FL, May 6-9.
Sign up by Dec. 31 to get the 2024 rate of just $495.
Register now!

Multiple Linear Regression in SAS

Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 3 replies
  • 1294 views
  • 0 likes
  • 2 in conversation