eyejay,
Thank you for your comments. It has been a couple of years since writing this article. Since then I have wanted to revisit this idea and fine tune the DMX queries. I will possibly do this with the release of SQL Server 2008.
Still have not hit the lottery, though.
Rodney

meganbearly 5/28/2008 5:56:26 AM

Excellent article, interesting cases are a good way to learn but as a bit of a novice, I was able to follow most of the text although a little more explanation of using the DMX query in Cluster mining would have been really helpful

Ian5/24/2008 5:27:11 PM

Hi Rodney, Thanks for the article. I am new to sql but, by doing these types of things, it really helps to broaden my understanding.

When I attempt to process the mining model, I get the following error message.

OLE DB error: OLE DB or ODBC error: The text data type cannot be selected as DISTINCT because it is not comparable.; 42000.

Thanks jjg

gray ghost,jjgonzales 6/13/2007 4:00:30 PM

I would take a look at the content types of the mining model fields making sure that the fields N1 to N6 are set to “discrete”. It might also be worthwhile to try different combinations of field types; the draw date will be the Key field and N1 to N6 can be Input Only or Input and Predict. In the DMX query, then, it is possible to use the number fields themselves in the predict join as well as the dates. For example, you might want to write a DMX query (or use the singleton query designer) that predicts the draw numbers for March, 24, 2007 but should return different results if the criteria was March, 24, 2007 AND N3 was 51 AND N2 was 21. Breaking out the DrawDate field to months, days and years, which I believe I did in the article, should also add more level of prediction.

I hope this helps. I will have to build the solution again in order to test these concepts as it has been quite a long time since I wrote the article. I certainly do appreciate the reader’s comments.

Thanks.
Rodney

Diana3/9/2007 12:48:26 PM

Hi Rodney,

I put together a mining package similar to the one you feature here but I can't get the query to return anything different for different input dates. It always returns the exact same set. I did get a different set for the clustering algorithm than I did for the decision tree. Any ideas as to how to modify the query so that the results reflect the day specified in the query?

Chris3/8/2007 11:25:02 AM