Association Rules

1) I use the identifier of transaction and attribute in one table.

Do I can to build a association rules structure without the use of the nested tables?

I tried - did not turn out...

2) As it is necessary to use a main and child table, can not build a prediction query.

When I try to add the predict column in a criteria/argument (Field=PredictSupport), i'll given message:

"Nested table column cannot be used as an argument in a data mining function."

I can not use other columns, because they are not predicable.

You have to have a nested table, but you can use the same table as the case and nested tables. Just mark the transaction key as the key for the "case" table, and the transaction item as the key for the "nested" table.

In your prediction query you would predict the nested table ( you are essentially predicting which rows will be there). For example, if your model was like this:

TransID LONG KEY,
Products TABLE PREDICT
(
ProductName TEXT KEY
)

You would issue a statement like

SELECT Predict(Products, 5) FROM MyModel PREDICTION JOIN ....

|||

Yes, turned out to build a model! Thanks!
But at an attempt to take advantage of builder of queries, there is an error:

Unable to retrieve the list of distinct states for mining model column Utp Doc Id.

ADDITIONAL INFORMATION:

Error (Data mining): The specified DMX column was not found in the context at line 1, column 17. (Microsoft SQL Server 2005 Analysis Services)

Without a builder - a query works :)
And yet question - as possible to apply a function, as though, PredictSupport? She needs a column, and for me does not turn out to use her from a table...

P.S.: I am sorry for my English =)