FORCE_REGRESSOR Error on Linear Regression Nested-Table Structure

Question

I have tried this structure in BIDS and as well as through DMX, still FORCE_REGRESSOR parameter is not working.  Its giving the following error:

Server: The operation has been cancelled.
Error (Data mining): 'HDD' is not a valid attribute display name.

My data mining structure is quite simple and looks like follows:

//DATA MINING STRUCTURE
CREATE MINING STRUCTURE Contracts_Structure_Nested
(
 [Customer]  LONG KEY
, Volumes   TABLE
 (
  [SiteOID] LONG KEY
 , [HDD]  LONG CONTINUOUS
 , [Volume] LONG CONTINUOUS
 )
)

 

//DATA MINING MODEL
ALTER MINING STRUCTURE Contracts_Structure_Nested
ADD MINING MODEL Contracts_Nested
(
 [Customer]
, Volumes
 (
  [SiteOID]
 , [HDD]  REGRESSOR
 , [Volume] PREDICT
 )
) USING MICROSOFT_LINEAR_REGRESSION (FORCE_REGRESSOR='{[HDD]}')

 

//POPULATE THE STRUCTURE WITH DATA
INSERT INTO Contracts_Nested
(
 Customer
, Volumes (SKIP, [SiteOID],[HDD],[Volume])
)
SHAPE
{
 OPENQUERY (
     ForDataMining
    , 'SELECT
        Customer
     FROM  dbo.Customer_Test
     ORDER BY Customer'
    )
}
APPEND
(
 {
  OPENQUERY (
      ForDataMining
     , 'SELECT
         Customer
      ,   SiteOID 
      ,   HDD
      ,   Volume
      FROM  dbo.Contracts_Test
      ORDER BY Customer'
     )    
 } RELATE Customer TO Customer
) AS Volumes

Any help will be highly appreciated.

Thanks,

Zafar

Answer 1

Hi Zafar,

I am working on this issue, and will give you an update soon.

Regards,

Jerry

TechNet Subscriber Supportin forum
If you have any feedback on our support, please contact  [e-mail]

Answer 2

Hello ZafarAlam, are you using Standard or Enterprise edition of SSAS?

Standard edition does not support this parameter. It will help if you can post build number of your SSAS.

Best regards, Vlad.

Answer 3

Here are the versions of my SQL server:

Microsoft SQL Server Management Studio      10.50.2500.0
Microsoft Analysis Services Client Tools      10.50.2500.0
Microsoft Data Access Components (MDAC)      6.1.7600.16385
Microsoft MSXML      3.0 6.0
Microsoft Internet Explorer      9.0.8112.16421
Microsoft .NET Framework      2.0.50727.4963
Operating System      6.1.7600

Thanks,

Zafar

Answer 4

Can you please execute SELECT @@Version on your SQL Server and check edition?

Thanks, Vlad.

Answer 5

Here's the result of the query:

Microsoft SQL Server 2008 R2 (SP1) - 10.50.2500.0 (X64)   Jun 17 2011 00:54:03   Copyright (c) Microsoft Corporation  Developer Edition (64-bit) on Windows NT 6.1 <X64> (Build 7600: )

Thanks,

Zafar

Answer 6

It should be compatible with Enterprise Edition by its features.

I need to take look on your complete project to make further suggestions - can you please open case with Microsoft Support to troubleshoot this issue?

Thank you, Vlad.

Answer 7

Vlad:

Support ticket/case is xxxxxxxxxxxxxx. I have got a reply from product support. I'm testing few things now.

Thanks,

Zafar

Answer 8

Hi Zafar, you might want to post solution to your issue here, I removed case details from your post.

Best regards, Vlad.

Answer 9

Vlad:

I'm working on it and very close. Will soon provide an update.

Thanks,

Zafar

Answer 10

Here is the answer that support provided to me:

The FORCE_REGRESSOR parameter is a little idiosyncratic from the user’s perspective because it takes attribute (display) names rather than column names, though it does makes sense based on how columns are mapped to mining attributes.

For case level attributes, this is straightforward because there is a 1:1 correspondence between columns and attributes. For nested tables, each key value is an attribute.  So, if you have a Beverages nested table,  your nested table attributes would be :  Beverages(Coke), Beverages(Pepsi), Beverages(Coffee), and so forth – one attribute for each member of the Beverages dimension (in OLAP-speak).

Attribute display names for nested table attributes are composed using unescaped names, so in the example below, you should specify something like

FORCE_REGRESSOR='{TestNTable(26).Yearly Income}' where 26 is one of the key values in the GeographyKey column.

See this article for an in-depth understanding of attributes vs. columns.

After all this, i tried the above tip in my model and it worked.