SS2005 Data Mining: Decision Tree implementation

Discussion in 'microsoft.public.sqlserver.datamining' started by Andriyev, Dec 2, 2009.

  1. Andriyev

    Andriyev Guest


    I have a SQL Server 2005 environment and there is this flat table
    representing my data. The attributes of the table are Gender, Ethnicity,
    Weather, GroupSize, Wait Time.

    Here all the attributes are discrete and the scenario is that of a person
    who intends to get space for his/her group at a restaurant.
    Ethnicity can take 4 values - American, African-American, Asian, Other
    Weather can take 4 values - Sunny, Windy, Rainy, Snow
    GroupSize can take 3 values - Less than 5, 5-10, More than 10
    Wait Time - none, 5-10mins., 11-30 mins., More than 30 mins

    Wait Time is the attribute to be predicted.

    Questions -

    1. Does the Decision Tree implementation in SQL Server 2005 and beyond
    support non-binary attributes i.e. attributes which can take more than 2
    values. Just like the ones mentioned above.

    I tried to create the decision tree mining model but all I get is a single
    node as part of the decision tree. Is there a limit on the values an
    attribute can take in the SQL Server implementation?

    2. I know its a little vague but is it possible to programmatically detect
    new attributes (which might get added in future to the database) and find the
    relevant ones in terms of the current model? In other words, how can the
    model evolve once its created?

    Thanks for your time.
    Andriyev, Dec 2, 2009
    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.