Forcasting Studio

[deleted_user]

hi all,
I need some help regarding the forecasting the demand.my data has two levels.
Headquater and resional office.Demand comes at all the facilities directly by coustomers.How can i forecast the demand at each facility using forecasting studio.

[udo_sas]

Hi -
SAS Forecast Studio will allow you to define these so-called hierarchical forecasting models in a very straightforward manner.
To illustrate how to organize your data let's assume that there are 2 locations A and B, your data starts in JAN05 up to DEC08.
You will need to organize your data similar to this:
TimeID Location Demand
01JAN05 A 100
01FEB05 A 200
01MAR05 A 200
... ... ...
01DEC08 A 100
01JAN05 B 100
01FEB05 B 200
01MAR05 B 200
...
01DEC08 B 100

When running the project wizard of SAS Forecast Studio you will need to specify "Location" as your hierarchy variable. By default SAS Forecast Studio will sum up the demands of each location to come up with the total demand at headquarter level. However, you can change this behavior to averaging, taking the max or min value, or similar.

For additional information please have a look at this whitepaper: http://www.sas.com/apps/whitepaper/index.jsp?cid=3495

I hope this helps.
Regards,
Udo

[deleted_user]

1 Thanks for clarification we have data like
TimeID Location Demand
01JAN05 A 100
01FEB05 A 200
01MAR05 A 200
... ... ...
01DEC08 A 100
01JAN05 B 100
01FEB05 B 200
01MAR05 B 200
...
01DEC08 B 100
now let we have 3 headquarters and 30 distribution centers.And different items are manufactured by different HQ's and distributed at distribution centers Further a costomer can order a item at HQ or distribution centers(DO) directly have to forecast the demand at each facility.Now suppose at any facility(DO) demand is intermittent and manufacturer want not to keep this item at that particular DO but item demand is shifted to HQ for that item.How can we cater this problem.
2.How I will check the forecasted accuracy of intermittent series,like for ARIMA MAPE value is measure for forecasting accuracy.

[udo_sas]

Hello -
Some more thoughts from my end - see below.
Regards,
Udo

> now let we have 3 headquarters and 30 distribution
> centers.And different items are manufactured by
> different HQ's and distributed at distribution
> centers Further a costomer can order a item at HQ or
> distribution centers(DO) directly have to forecast
> the demand at each facility.
In your case setting up so-called hierarchical model seems not to be possible - as you are facing (stochastic) demand on every level of the hierarchy (opposed to facing demand only on the lowest level or facing deterministic demand - or known demand - on higher levels). Instead you will need to forecast each level in a separate project. Note that this is possible in Forecast Server, too. You will need to split your data set accordingly. You can the TIMESERIES procedure (or the respective EG task) to aggregate data to the respective level.

If there is a need to reconcile all these different projects to make sure the predictions add up Forecasts Servers provides you with some means to archive this goal as well.

> Now suppose at any facility(DO) demand is intermittent and manufacturer
> want not to keep this item at that particular DO but
> item demand is shifted to HQ for that item.How can we
> cater this problem.
As far as I can tell this problem is not only a forecasting problem, but an inventory replenishment (or inventory optimization) problem actually. You might want to check out these materials:
http://www.sas.com/apps/whitepaper/index.jsp?cid=3520
http://www.sas.com/industry/mfg/iobrochure.pdf

> 2.How I will check the forecasted accuracy of
> intermittent series,like for ARIMA MAPE value is
> measure for forecasting accuracy.
You might be interested in this paper: http://support.sas.com/resources/papers/sgf2008/forecasting.pdf which outlines how SAS Forecast Server handles intermittent demand series.

[deleted_user]

Dear Sir Thanks for your suggestions,Now I have some questions regarding Forecast Studio ,that What should be the value of these parameters in SAS Forecasting Studio:
1.Perform intermittency test. sensitivity
2.Perform seasonality test. sensitivity
3.Outlier detection-detect outliers
4.Maximum percentage of series
5.forecast studio only detect outliers in case of ARIMA?
6. What is the meaning of significance level in outlier detection
7.how can we use outliers for our forecasting.