none
Interestingness score & Shannon's Entropy in Analysis Services Server (Clustering)

    Question

  • Can Analysis Services Server provide Interestingness & Shannon's scores? How can I get these values per model attribute\feature (problematically or using a mdx query)?

    See also: http://msdn.microsoft.com/en-us/library/ms175382.aspx

    Thanks!

    • Moved by Deepak Puri Sunday, July 18, 2010 2:29 AM Question on Data Mining Algorithms (From:SQL Server Analysis Services)
    Saturday, July 17, 2010 4:27 PM

Answers

  • No, you cannot extract the entropy scores. Most algorithms support a MAXIMUM_INPUT_ATTRIBUTES parameter which drives the feature selection, which is based on entropy, but the individual scores are not exposed.

    Probably the closest you can get to interestingness ranking (using the built-in features) is to build a Naive Bayes model and rank the features based on the attribute discrimination stored procedures.


    bogdan crivat / http://www.bogdancrivat.net/dm
    Friday, July 23, 2010 4:48 AM