Analytic Functions for information on the syntax, semantics, and restrictions of
CLUSTER_PROBABILITY returns a probability for each row in the selection. The probability refers to the highest probability cluster or to the specified
cluster_id. The cluster probability is returned as
CLUSTER_PROBABILITY can score the data in one of two ways: It can apply a mining model object to the data, or it can dynamically mine the data by executing an analytic clause that builds and applies one or more transient mining models. Choose Syntax or Analytic Syntax:
Syntax — Use the first syntax to score the data with a pre-defined model. Supply the name of a clustering model.
Analytic Syntax — Use the analytic syntax to score the data without a pre-defined model. Include
nis the number of clusters to compute, and mining_analytic_clause, which specifies if the data should be partitioned for multiple model builds. The
order_by_clause. (See analytic_clause::=.)
The syntax of the
CLUSTER_PROBABILITY function can use an optional
GROUPING hint when scoring a partitioned model. See GROUPING Hint.
mining_attribute_clause identifies the column attributes to use as predictors for scoring. When the function is invoked with the analytic syntax, these predictors are also used for building the transient models. The
mining_attribute_clause behaves as described for the
PREDICTION function. (See mining_attribute_clause::=.)
The following example is excerpted from the Data Mining sample programs. For more information about the sample programs, see Appendix A in Oracle Data Mining User’s Guide.
The following example lists the ten most representative customers, based on likelihood, of cluster 2.
SELECT cust_id FROM (SELECT cust_id,
OVER(ORDER BY prob DESC, cust_id) rnk_clus2 FROM (SELECT cust_id, CLUSTER_PROBABILITY(km_sh_clus_sample, 2 USING *) prob FROM mining_data_apply_v)) WHERE rnk_clus2 <= 10 ORDER BY rnk_clus2; CUST_ID ---------- 100256 100988 100889 101086 101215 100390 100985 101026 100601 100672