Attribute Limit of 5000 When Mining a Cube

I'm trying to build a association model in the Standard Edition based on an existing cube. I keep getting the error:

Error (Data mining): The 'Product Recommendations' mining model has 60385 attributes. This number of attributes exceeds the attribute limit of 5000 allowed by the current version of the algorithm associated with the mining model.

I created Cube Slice filters and those limit the Customer and Product dimensions (Product is Nested) to well under 5000. The error message also does not change. The number of attributes is equal to the number of rows in the Product dimension, but I expected the cube slice to reduce the number. I tested all the SQL used while it processes and with the MDXFilters the number of rows returned is well under 5000.

So, in short, the final questions is, is it possible to create a mining model in standard edition based on an existing cube where the nested dimension in the model has more than 5000 rows? Is there some other way to filter the query?

I guess my only choice on this if there isn't a way is to extract the data into relational table with only the rows I want to analyze....that's a huge pain and doesn't really make sense when the filters should limit the model size.

What is annoying on this is I can't find one reference anywhere on the microsoft site that this limit even exists within the product...

Thanks for any assistance on this!

This sounds like a bug (and knowing the code, I can see where it comes from).

Standard edition is limited to 5000 attributes for Standard Edition. What you are doing is perfectly legal, assuming there are actually less than 5000 attributes. You may need to contact support so we can get this fixed, or start by writing up a bug report at connect.microsoft.com. If you cannot get this resolved through support and this is blocking you, please let us know and we will see if we can make something happen.

Thanks

-Jamie

|||

Thanks...I'll get a support case open and see what happens.

- L

|||

Jamie,

I opened a support case with Microsoft and have been working with them for a few days. They did agree that this is a defect in the product. The current response is that, "they may consider a design change to account for filtered cube slices". This is a pretty big deal for our development. I'd also question the limit in general since its not even documented, does the limit really drive more sales to the Enterprise Version of the product? I think if there is a limit, it should really be much larger since in practice, 5000 rows in a nested table is so small. It is very common for small companies to have product catalogs of 60k products.

THANKS!

|||

I have been working with the support person you are in contact with. The issue isn't if the problem with be fixed, it is simply how the problem will be fixed. I believe we will have a resolution today - in any case, it will resolve the issue you are seeing.

Regarding the 5000 attribute limit, that's a completely seperate debate that I don't want to get deeply into in the forums. I would only state that the differentiation between the various editions isn't limited to data volumes, functional differences arise as well (e.g. data mining transforms in Integration Services, etc.). Your cost structure in general may not support a wholesale changeover to Enterprise edition due to data volumes, but the ROI of some of the features that are made available in the larger edition may warrant an upgrade on a single machine. There are different licensing options for this as well, i.e. server CAL vs Processor licences, which for a small-medium size business, a 50-CAL Enterprise license may do the trick and still be in the affordable price range.

If you want to contact me directly, it's Jamie <dot> MacLennan <at> microsoft <dot> com

Thanks