cancel
Showing results for 
Search instead for 
Did you mean: 

Regarding MLR model and How it is configured?

former_member309586
Participant
0 Kudos

Hi Gurus,

I have one requriment which client wanted me to forecast indexes which is cost factor.

My question is how to forecast the indexes which is dependent on external factors like product values?

I think MLR will help me and i dont know how to configure it.

If any one can help me it will be great help.

Thanks

Raj

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Hi Raj,

You can use MLR model , but remember that MLR is purely mathematical and the results will be not be that satisfactory , unless the depenedent variables are fully explained by the independent variables.

The general MLR equation as you are aware

Y = b 0 + b 1X1 + b 2X2 + b 3X3...b nXn + ei

Where: Y = Dependent variable , b 0 = Y intercept or constant, b i = Coefficients or weights, X i = Independent variables, ei = Residual or prediction error.

Here in your situation Y is the forecast , X1 is the cost factor , there may be other factors that influence the forecast which can be termed as X2 , X3 etc . MLR needs historical data of not only the Sales order , but also as that of the cost factor and other variables . This is because based on the pattern of sales history & independent key figures MLR will derive an equation as specified above ( This will not been in the screen , we need to interpret after seeing the result in the planning book) . From the equation we can interpret whether independent variables do really have a effect on the dependent variable.

Once you have data for all the Key figures , go & create a MLR profile in maintain forecast profile . go to MLR tab and enter your dependent Key figure which is historical sales & independent variables which may key figures or time series . Save the profile.

Now go to the planning book which has MLR tab . Run the forecast and see the result . Change the independent variables & check how it is affecting the Forecast .

The results may not be good , as MLR is purely a mathematical equation & the coeffients which you will get may be negative (sometimes) , which business users might not accept .

Please do the testing & let me know in case of further clarifications.

Regards

Surendra

Answers (1)

Answers (1)

Former Member
0 Kudos

Not sure what you mean by 'Indexes' and how MLR will help you in forecasting it.

If you need to configure a MLR profile then you need to decide which KFs in your planning area will be the 'dependent' variables and which ones would be the 'independent' variables. For ex: if you Sales is a function of Advertising then Sales is dependent & advertising is independent. The MLR relation is represented by

Y = A + BX; where Sales = Y & Advt = X

You need to define these key figures in your profile and the system uses these to calculate the constants A,B, .... when you execute the forecast. there are a number of other things like Diagnosis Group etc ..and suggest you use F1 to study more about it.

The key factor is defining your 'dependent' & 'independent' variables such that there actually exists a co-relation between them. Hope this helps.

-Srinidhi