Microsoft Developer Network > Forenhomepage > Data Mining > What is the Excel Addin Doing
Stellen Sie eine FrageStellen Sie eine Frage
 

BeantwortetWhat is the Excel Addin Doing

  • Freitag, 12. Juni 2009 11:27Matt Tolhurst TeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillen
     
    Hi,
    I am just starting to use the excel add in and some of the things I can't figure out.  I am trying to do forecasting, which every demo i have seen looks simple enough.  i have created a set of data using two facts/measures and the first of every month for 2 years.  When i use the table tools, Analyze, Forecast I get different results depending on how i do it.

    e.g.
    My data (some of)
    01/03/2009 00:00 3583973.49 2969
    01/04/2009 00:00 3057092.23 2519
    01/05/2009 00:00 2053462.79 2382

    if i forecast both measures at the same time
    01/06/2009 00:00 3948649.77 3014
    01/07/2009 00:00 4175897.669 3382
    01/08/2009 00:00 3948649.77 3296
       
       

    if i forecast them on their own

    first measure
    3992107
    3992107
    3992107
    3992107
    3992107

    second measure

    2776
    2988
    2631
    2708
    2476

    Can anyone explain why it is doing that?  I am guessing that when you do it together they both impact one another, can i get it so that measures don't influence others?  And why do I only get 3? As opposed 5 when i do them seperately?

    A very confused
    Matt

Antworten

  • Montag, 29. Juni 2009 15:25Allan MitchellMVPTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillen
     Beantwortet
    Matt

    Which version of SQL Server are you using?  2005?

    If you are using 2005 then the algorithm used for all time series predictions is ARTxP.  The x is "Cross" and this allows a variable to be influenced by another.  This is very probably the reason you are seeing the difference in values.

    The second Q which is "Why does it stop".  Again in 2005 what happens is that if the predictions start to get too wild then SSAS will just stop predicting and you have no way of controlling it.  ARTxP is great at "Short Term/Next Value" prediction but not as good further out.  That is one of the reasons why ARIMA was introduced in 2008.


    Allan

Alle Antworten

  • Montag, 29. Juni 2009 15:25Allan MitchellMVPTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillen
     Beantwortet
    Matt

    Which version of SQL Server are you using?  2005?

    If you are using 2005 then the algorithm used for all time series predictions is ARTxP.  The x is "Cross" and this allows a variable to be influenced by another.  This is very probably the reason you are seeing the difference in values.

    The second Q which is "Why does it stop".  Again in 2005 what happens is that if the predictions start to get too wild then SSAS will just stop predicting and you have no way of controlling it.  ARTxP is great at "Short Term/Next Value" prediction but not as good further out.  That is one of the reasons why ARIMA was introduced in 2008.


    Allan

  • Montag, 29. Juni 2009 19:24Matt Tolhurst TeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillenTeilnehmermedaillen
     
    Thanks Allan,

    We are using 2005, there is currently some arm twisting going on so we can get 2008.  I will add what you say to my case :)