Data Mining Time Series Strange Forecast

Discussion in 'microsoft.public.sqlserver.datamining' started by Val Matison, May 2, 2005.

  1. Val Matison

    Val Matison Guest

    Hi,

    I'm going through the tutorials for Data Mining and have issues with the
    time series example. I built the model and started looking at some of the
    forecasts. Some of them looked plausible, others like R750 North America
    Amount are so far off that it makes the entire model suspect. In two steps
    the amount skyrockets to 94,389 from a peak of 38,339 as the last actual
    value. There is nothing in the history of that model that would support such
    a huge spike.

    Can someone please tell me I'm reading this incorrectly or is there a
    problem with this algorithm?

    BTW, I'm using the latest CTP.

    Thanks
     
    Val Matison, May 2, 2005
    #1
    1. Advertisements

  2. There are some known issues where the TS algo starts to osciallate that we
    have fixed since the last CTP build and are continually tuning the
    algorithms. You should check the deviation along with the predictions -if
    the deviations start getting too high, then the algorithm is telling you it
    is getting unsure of the results.
     
    Jamie MacLennan \(MS\), May 3, 2005
    #2
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.