Dela via


SELECT INTO (DMX)

Creates a new mining model that is built on the mining structure of an existing mining model. The SELECT INTO statement creates the new mining model by copying schema and other information that is not specific to the actual algorithm.

Syntax

SELECT INTO <new model> 
USING <algorithm> [(<parameter list>)] [WITH DRILLTHROUGH[,] [FILTER(<expression>)]]
FROM <existing model>

Arguments

  • new model
    A unique name for the new model that is being created.

  • algorithm
    The provider-defined name of a data mining algorithm.

  • parameter list
    Optional. A comma-separated list of provider-defined parameters for the algorithm.

  • expression
    An expression that evaluates to a valid filter condition on the training data. For more information about expressions that can be used as filters, see Creating Filters for Mining Models (Analysis Services - Data Mining).

  • existing model
    The name of the existing model to be copied.

Remarks

If the existing model is trained, the new model is automatically processed when this statement executes. Otherwise, the new model remains unprocessed.

The SELECT INTO statement works only if the structure of the existing model is compatible with the algorithm of the new model. Therefore, this statement is most useful for rapidly creating and testing models that are based on the same algorithm. If you change the algorithm type, the new algorithm must support the data type of each column that is in the existing model, or an error might occur when the model is processed,

The WITH DRILLTHROUGH clause enables drillthrough on the new mining model. Drillthrough can only be enabled when you create the model.

Example 1: Altering the Parameters of the Model

The following example creates a new mining model based on an existing mining model, TM_Clustering, which you create in the Basic Data Mining Tutorial. In the new model, the CLUSTER_COUNT parameter is modified so that a maximum of five clusters will exist in the new model. In contrast, the existing model uses the default value, which is 10.

SELECT * INTO [New_Clustering]
USING [Microsoft_Clustering] (CLUSTER_COUNT = 5) 
FROM [TM Clustering]

Example 2: Adding a Filter to the Model

The following example creates a new mining model based on an existing mining model, and adds a filter on the model. The filter restricts the training data to only those customers who live in a particular region.

SELECT * INTO [Clustering Europe Region]
USING [Microsoft_Clustering] WITH FILTER(Region='Europe')
FROM [TM Clustering]

Note

Filters that are applied to the case table can be altered by using the SELECT INTO statement as shown in this example; however, if the original model contains a filter on a nested table, the nested table filter cannot be altered or removed by using this syntax, but is copied unchanged from the original model. To create a model with a different filter on a nested table, use the ALTER STRTUCTURE...ADD MODEL syntax.