TM1 Forum

Can we do trending in TM1 like based on last 5 year data by using one of the technique

Moving Averages.
Weieghted Moving Averages
Exponential Smoothing. etc.

Basically, the answer is yes for the first two; I have done this before. The third sounds like a type of linear regression and this is also possible. As to how, you need to dig into c-level rules. Your time dimensions are going to be an important factor in how difficult it will be to model these trends; as will the number of dimensions in the cube you are trying to do the calculations in.

rmackenzie wrote:Basically, the answer is yes for the first two; I have done this before. The third sounds like a type of linear regression and this is also possible. As to how, you need to dig into c-level rules. Your time dimensions are going to be an important factor in how difficult it will be to model these trends; as will the number of dimensions in the cube you are trying to do the calculations in.

Hi,

I have a sales cube having these dimensions.
Year
Month
Facility
Customer Class
Product Line
Market
Item Sales Group
Scenario
Sales Measures

And i want to get the trend using the last 5 years data , I tried doing it using rules but the cube was running very slow, so what I did is I calculated the average in SQL database and stored it in a table and loaded it from TI Process.

However at consolidations it is giving me the sum of it, which is actually not the average. Now I have to go back to rules again to do calculation for the consolidation levels.

Can you give me a hint or a starting point. I will appreciate it. Thanks.

There's not just one way to do averages - which way will depend, in part, on how your measures are set-up in 'Sales Measures'. If your performance is poor then you need to consider the cube design - do you need to know the averages split by all the other dimensions; are you feeding correctly; can you use feeder-less rules?

Try this for a start (just the IBM doco).