http://social.technet.microsoft.com/wiki/contents/articles/7566.a-guide-to-the-undocumented-system-stored-procedures-for-data-mining.aspx
These data mining stored procedures are officially undocumented and unsupported but nonetheless very handy.

Sequence Clustering Models

Most of the stored procedures used for sequence clustering models have the same names as those used for clustering models; however, the number of arguments can be different and the results from a sequence clustering model include transitions and states as well as any other attributes that might have been used for clustering.

This section shows only examples for models based on the Microsoft Sequence Clustering algorithm.

GetClusterCharacteristics

Returns the states and transitions in the specified cluster.

Fully qualified procedure name

System.Microsoft.AnalysisServices.System.DataMining.Clustering.GetClusterCharacteristics

Parameters

Model name as string

Node ID as string (optional)

Threshold value as double (optional)

Example

The example returns all the characteristics of Cluster 12 that have a probability score over the specified threshold of 0.05.

CALL System.GetClusterCharacteristics('Sequence Clustering','12',0.05) 

Sample results

Attributes	Values	Frequency	Support
Model.Transitions	[Start] -> Touring-1000	0.135565537	285
Model	Water Bottle	0.119103087	Mt;width:25%;background-color:transparent;"> 251
Model.Transitions	[Start] -> Mountain-200	0.093507554	197
Model	Sport-100	0.085614444	180

 

Notes

This stored procedure is used by the Microsoft Sequence Clustering Viewer.

See also

Queries on a Sequence Clustering Model (MSDN)

 

GetClusterDiscrimination

Returns the attributes that are most different between the two specified clusters, or that distinguish one cluster from the rest of the cases in the model.

Fully qualified procedure name

System.Microsoft.AnalysisServices.System.DataMining.Clustering.GetClusterDiscrimination

Parameters

Model name as string

Node unique id as string

Second node unique id as string, or omit for complement

Probability threshold as double

Rescaled as Boolean

Example 1

The following example compares the characteristics of Cluster 9 with its complement, but only returns attributes for which the probability score is over the specified threshold of 0.005.

The rescaled argument true indicates that the probabilities should be normalized.

CALL System.GetClusterDiscrimination('Sequence Clustering','009','',0.8, true)

Example 2

The following example compares the characteristics of Cluster 1 and Cluster 12, but returns only attributes that have an absolute score greater than 10.

CALL System.GetClusterDiscrimination('Sequence Clustering','1','12',10,true)

A positive value for Scores indicates that the attribute favors the first cluster (1); a negative score indicates that the attribute favors the second cluster (12).

If a second node is not specified for comparison, the procedure returns the complement of the first node, meaning all remaining cases not in the first specified node. If you specify a node ID of 0, the attributes for the entire model are returned.

Sample results 2

Attributes	Values	Score
Income Group	Low	100
Score
Income Group	Income Group	High	-90
Model.Transitions	Mountain-400-W-> [End]	-48
Model.Transitions	LL Mountain Tire->[End]	-45
Model.Transitions	Touring Tire Tube->[End]	-36

 

Notes
This stored procedure is used by the Microsoft Sequence Clustering Viewer. 

See also

Queries on a Sequence Clustering Model (MSDN)

GetClusterProfiles
Returns the characteristics of the cluster, with all the information that is found in the NODE_DISTRIBUTION table for the cluster.

Fully qualified procedure name

System.Microsoft.AnalysisServices.System.DataMining.Clustering.GetClusterProfiles

Parameters

Model name as string

Count of attributes as integer

ID of attribute to start at as integer

Example

This example returns all of the clusters in the model, with their respective profiles. Each profile shows the distribution of attribute values for the attributes specified by the second and third arguments.

CALL System.GetClusterProfiles('Sequence Clustering', 1, 0) 

ID of attribute to start at as integer

Sample results

AttributeName	AttributeValue	AttributeValueType	ValueTag	Marginal	1
Model	Touring-1000	1	Probability	0	2.22021
Model	Touring-1000	1	Support	0	0.0003
Model	Touring-2000	4	Probability	0.286847234357148	0.846601
Model	Touring-2000	4	Support	4268	1487.1309
Model	Touring-3000	4	Probability	0.491699	2.1685
Model	Touring-3000	4	Support	7316	0.00038

 

Notes

This stored procedure executes a series of DMX statements that return the complete set of profiles in the model. The clusters are not returned in any specific order and you cannot restrict the results by cluster.

See also

Queries on a Sequence Clustering Model (MSDN)

 

 

GetClusters

 

Returns a list of all clusters in the model, with the node ID and cluster name.

Fully qualified procedure name

System.Microsoft.AnalysisServices.System.DataMining.Clustering.GetClusters

Parameters

Model name as string

Example

This example gets a list of all clusters in the model, with the ID of each cluster. If you have renamed the cluster, the new name is returned.

CALL System.GetClusters('SequenceClustering')

Sample results

ClusterName	ClusterUniqueNodeID
Renamed cluster 1	1
Cluster 2	2

 

Notes

None

See also

Queries on a Sequence Clustering Model (MSDN)

Return to list of stored procedures for all model types