BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
baimeeker
Calcite | Level 5

I am working on a forecast value add project where I plan to compare SAS forecasts for multiple SKUs to our current forecasting system. To do this, I will need to export forecasts for each SKU into a data set that I can compare against our current forecasts.  This works fine for SKUs with continuous demand, but in order to compare the forecasts for intermittent demand I will need forecasts for both the Average Demand Size and the Demand Interval for these SKUs.  I can export these by SKU, however I'd like to design a data set of forecasts for multiple SKUs.  How do I produce the Demand Interval forecasts using the data sets exported from HPF?  I can see parameters in OUTEST, but I do not know how to fit those together to get a number.

 

I have been using the following code and I am attaching an image to display the ODS output I receive that shows me the interval estimate.

 

The TL/DR version is: How does ODS calculate the Demand Interval Prediction?

 

 * assign project settings ;
         %let HPF_SETMISSING=0;
         %let HPF_ZEROMISS=NONE;
         %let ODSSTYLE=STATISTICAL;
         %let HPF_TRIMMISS=BOTH;
         %let HPF_BACK=0;
         %let HPF_START='07Apr2014'd;
         %let HPF_START_ENABLED=0;
         %let HPF_END='14Mar2016'd;
         %let HPF_END_ENABLED=0;
         %let HPF_HORIZON_START='21Mar2016'd;
         %let HPF_HORIZON_START_ENABLED=0;
         %let HPF_SELECT_MINOBS_TREND=2;
         %let HPF_SELECT_MINOBS_SEASONAL=2;
         %let HPF_SELECT_MINOBS_NON_MEAN=2;
         %let HPF_LEAD=4;
         %let HPF_DIAGNOSE_INTERMITTENT=2.0;
        %let HPF_SELECT_HOLDOUT=4;
         %let HPF_SELECT_HOLDOUTPCT=50;
         %let HPF_SELECT_ENDZEROS_MAXNUM=5;
         %let HPF_SELECT_ENDZEROS_MAXPCT=50;
         %let HPF_SELECT_ENDZEROS_MINOBS=1;
         %let HPF_SELECT_CRITERION=MAE;
         %let HPF_COMPONENTS=INTEGRATE;
         %let HPF_FORECAST_ALPHA=0.05;
         %let HPF_RECONCILE_WEIGHTED=;
         %let HPF_RECONCILE_IGNOREMISSF=;
         %let HPF_RECONCILE_LOCKZERO=;
         %let HPF_RESTRICT_OVRD_RECON_DIR=0;
         %let HPF_FORCE_OVRD_CONFLICT_CHECK=1;
         %let HPF_DO_NOT_ALLOW_NEG_FORECAST=;
         %let HPF_CREATE_INDEX_FILES=;
         %let HPF_CREATE_OUTCOMPONENT_DATA_SET=;
         %let HPF_DIAGNOSE_ARIMAX=;
         %let HPF_DIAGNOSE_ESM=;
         %let HPF_RETAINCHOOSE=RETAINCHOOSE;
         %let HPF_SELECTION_LIST=FSDEMO.CUST_REPO.TOP_2;
         %let HPF_OUTLIER_FILTER=SUBSET;
         %let HPF_COMBINE=;
         %let HPF_COMBINE_CRITERION=RMSE;
         %let HPF_COMBINE_MISSMODE=MISSING;
         %let HPF_COMBINE_METHOD=AVERAGE;
         %let HPF_COMBINE_STDERR=DIAG;
         %let HPF_ARIMA_IDENTIFY=BOTH;
         %let HPF_USE_DIAGNOSE=;
         %let HPF_TRANSTYPE=NONE;
         %let HPF_TRANSOPT=MEDIAN;
 proc hpfengine data=_leaf.DATA (where= (corp_item_num=4756656.0 or corp_item_num= 2397586.0))
        inest=_leaf.outest
		seasonality=52
        errorcontrol=(severity=ALL, stage=(PROCEDURELEVEL DATAPREP SELECTION ESTIMATION FORECASTING))
        EXCEPTIONS=CATCH modelrepository=work.TemLevModRep_leaf      
		print=(FORECASTS select estimates)
        task = select( alpha=&HPF_FORECAST_ALPHA criterion=&HPF_SELECT_CRITERION endzeros=(maxnum=&HPF_SELECT_ENDZEROS_MAXNUM) endzeros=(maxpct=&HPF_SELECT_ENDZEROS_MAXPCT) endzeros=(minobs=&HPF_SELECT_ENDZEROS_MINOBS) holdout=&HPF_SELECT_HOLDOUT holdoutpct=&HPF_SELECT_HOLDOUTPCT minobs=&HPF_SELECT_MINOBS_NON_MEAN  minobs=(season=&HPF_SELECT_MINOBS_SEASONAL) minobs=(trend=&HPF_SELECT_MINOBS_TREND)
	
        seasontest=none intermittent=&HPF_DIAGNOSE_INTERMITTENT override)
        back=&HPF_BACK components=&HPF_COMPONENTS lead=&HPF_LEAD
       out=work.out outfor=work.outfor outstat=work.outstat
	   outest=work.outest
        outstatselect=work.outstatselect
        outmodelinfo=work.outmodelinfo
       scorerepository=work.scorerepository outsum=work.outsum
        outcomponent=work.outcomponent
       inevent=_project.EventRepository;
        by client corp_item_num ;
		score;
       id week_of interval=WEEK.2 format=YYMMDD10. acc=total notsorted horizonstart=&HPF_HORIZON_START start=&HPF_START;
       forecast rawdmd /
         setmissing=&HPF_SETMISSING trimmiss=&HPF_TRIMMISS zeromiss=&HPF_ZEROMISS
       ;
	   
       run;

I'm using SAS 9.3 and EG 5.1.  This is all using Forecast Server 12.1.

 


2016-04-20 10_49_45-SAS Enterprise Guide - Project3.png
1 ACCEPTED SOLUTION

Accepted Solutions
alexchien
Pyrite | Level 9

The interval and size forecasts are stored in the ODS table called Demands. You can "capture" the ODS table using ODS OUTPUT Demands = idm_demand_predictions; and change the PROC HPFENGINE option PRINT= option to PRINT = ALL. If you don't want all the ODS ouptuts printed in the listing/html output, you can use the following trick before and after the proc call to capture just the Demand ODS table:

 

ods exclude all;
ods output demands = idm_demand_predictions;

 

proc hpfengine ...... print = all;
 ......
run;
ods exclude none;

View solution in original post

2 REPLIES 2
alexchien
Pyrite | Level 9

The interval and size forecasts are stored in the ODS table called Demands. You can "capture" the ODS table using ODS OUTPUT Demands = idm_demand_predictions; and change the PROC HPFENGINE option PRINT= option to PRINT = ALL. If you don't want all the ODS ouptuts printed in the listing/html output, you can use the following trick before and after the proc call to capture just the Demand ODS table:

 

ods exclude all;
ods output demands = idm_demand_predictions;

 

proc hpfengine ...... print = all;
 ......
run;
ods exclude none;

baimeeker
Calcite | Level 5
Thank you so much! This gives me just what I was looking for.

SAS Innovate 2025: Save the Date

 SAS Innovate 2025 is scheduled for May 6-9 in Orlando, FL. Sign up to be first to learn about the agenda and registration!

Save the date!

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
  • 2 replies
  • 1272 views
  • 0 likes
  • 2 in conversation