Composite Profile

[Former Member]

I am attempting to get some clarity on the logic of the Composite Profile. In the Foti Chimni text entiteled "Demand Management with SAP". part of the composite definition goes on to state that the underlying objective is to take advantage of the strenghts of each model to create a single one number forecast. either by averaging the forecats, giving each one equal weight , or one can weight each one differently. My company was under the assumption that the composite model will choose the "best fit univariate profile" based on lowest MAPE ect. Is this assumption correct? I just want to clarify what SAP means by averaging the forecast if equal weight is given to each univariate model within the Master composite profile

[Former Member]

Melissa,

A composite profile is atricky business. It is probbaly best used in a business which has multiple 'best-fit' scenarios (i know it sounds like an oxymoron!). The idea is to take some from all. Something which is very rarely used. If you dont have any external influences to your business and causal factors are ruled out, I would suggest its best to go by the Univariate models only.

"My company was under the assumption that the composite model will choose the "best fit univariate profile" based on lowest MAPE ect. Is this assumption correct?"

I am afraid not. If the intent is to come up with a univariate model with the least amount of error why dont you try and use "Automatic Model Selection procedure 2" Startegy 56?

http://help.sap.com/saphelp_scm50/helpdata/en/8f/9d6937089c2556e10000009b38f889/frameset.htm

You can give the error types and ranges in the diagnostic group.

Cheers!

Abhi