time series historic_model_count and gap????

Discussion in 'microsoft.public.sqlserver.datamining' started by aricka, Mar 25, 2006.

  1. aricka

    aricka Guest

    if i have a model that has arrivals spreading on two years aggregated by an hour meaning 24 points a day and 24*7 a week. what would be the correct parameters for the time series historic_model_count, gap and hint????
     
    aricka, Mar 25, 2006
    #1
    1. Advertisements

  2. There are no "correct" values for historic_model_count and _gap, since they
    exist only for validation purposes. If you wanted to train daily and see
    how a previous days model predicted for the following day, I would use 1 and
    24. If you wanted to train weekly and see how well a model would predict
    for an entire week, I would use 1 and 168.

    For PERIODICITY_HINT, it seems that there are obvious periodicities at the
    daily and weekly levels, so 24 and 168 would work as well.

    --

    -Jamie MacLennan
    SQL Server Data Mining
    This posting is provided "AS IS" with no warranties, and confers no rights.
    PLEASE POST NEW QUESTIONS AT
    http://forums.microsoft.com/msdn/showforum.aspx?forumid=81&siteid=1
     
    Jamie MacLennan \(MS\), Mar 27, 2006
    #2
    1. Advertisements

  3. aricka

    aricka Guest

    thanks for the answer i will check it with my model, another quesion that i am wondering about is, in the time series model if want to do predictions for each hour for the comming year and use it in a simulation software, how can i cache the data that the predict function would return an immediate answer and not take a long time?
     
    aricka, Mar 28, 2006
    #3
    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.