SystemGetClusterCrossValidationResults (Analysis Services - Data Mining)

Applies to: SQL Server 2019 and earlier Analysis Services Azure Analysis Services Fabric/Power BI Premium

Partitions the mining structure into the specified number of cross-sections, trains a model for each partition, and then returns accuracy metrics for each partition.

Note This stored procedure can be used only with a mining structure that contains at least one clustering model. To cross-validate non-clustering models, you must use SystemGetCrossValidationResults (Analysis Services - Data Mining).

Syntax

  
SystemGetClusterCrossValidationResults(  
<structure name>,   
[,<mining model list>]  
,<fold count>}  
,<max cases>  
<test list>])  

Arguments

mining structure
Name of a mining structure in the current database.

(required)

mining model list
Comma-separated list of mining models to validate.

If a list of mining models is not specified, cross-validation is performed against all clustering models that are associated with the specified structure.

(optional)

fold count
Integer that specifies the number of partitions into which to separate the data set. The minimum value is 2. The maximum number of folds is maximum integer or the number of cases, whichever is lower.

Each partition will contain roughly this number of cases: max cases/fold count.

There is no default value.

(required)

max cases
Integer that specifies the maximum number of cases that can be tested.

A value of 0 indicates that all the cases in the data source will be used.

If you specify a number that is higher than the actual number of cases in the data set, all cases in the data source will be used.

(required)

test list
A string that specifies testing options.

Note This parameter is reserved for future use.

(optional)

Return Type

The Return Type table contains scores for each individual partition and aggregates for all models.

The following table describes the columns returned.

Remarks

To return accuracy metrics for the entire data set, use SystemGetClusterAccuracyResults (Analysis Services - Data Mining).

Also, if the mining model has already been partitioned into folds, you can bypass processing and return only the results of cross-validation by using SystemGetClusterAccuracyResults (Analysis Services - Data Mining).

Examples

The following example demonstrates how to partition a mining structure into three folds, and then test two clustering models that are associated with the mining structure.

Line three of the code lists the specific mining models that you want to test. If you do not specify the list, all clustering models associated with the structure are used.

Line four of the code specifies the number of folds, and line five specifies the maximum number of cases to use.

Because these are clustering models, you do not need to specify a predictable attribute or value.

CALL SystemGetClusterCrossValidationResults(  
[v Target Mail],  
[Cluster 1], [Cluster 2],  
3,  
10000  
)  

Sample results:

Requirements

Cross-validation is available only in SQL Server Enterprise beginning in SQL Server 2008.

See Also

SystemGetCrossValidationResults (Analysis Services - Data Mining)
SystemGetAccuracyResults (Analysis Services - Data Mining)
SystemGetClusterCrossValidationResults
SystemGetClusterAccuracyResults (Analysis Services - Data Mining)