44.17 PREDICTION_SET

Syntax

prediction_set::=

Analytic Syntax

prediction_set_analytic::=

cost_matrix_clause::=

mining_attribute_clause::=

mining_analytic_clause::-

See Also:

"Analytic Functions" for information on the syntax, semantics, and restrictions of mining_analytic_clause

Purpose

PREDICTION_SET returns a set of predictions with either probabilities or costs for each row in the selection. The return value is a varray of objects with field names PREDICTION_ID and PROBABILITY or COST. The data type of the PREDICTION field depends on the target value type used during the build of the model; the probability and cost fields are BINARY_DOUBLE.

PREDICTION_SET can perform classification or anomaly detection. For classification, the return value refers to a predicted target class. For anomaly detection, the return value refers to a classification of 1 (for typical rows) or 0 (for anomalous rows).

bestN and cutoff

You can specify bestN and cutoff to limit the number of predictions returned by the function. By default, both bestN and cutoff are null and all predictions are returned.

  • bestN is the N predictions that are either the most probable or the least costly. If multiple predictions share the Nth probability or cost, then the function chooses one of them.

  • cutoff is a value threshold. Only predictions with probability greater than or equal to cutoff, or with cost less than or equal to cutoff, are returned. To filter by cutoff only, specify NULL for bestN. If the function uses a cost_matrix_clause with COST MODEL AUTO, then cutoff is ignored.

You can specify bestN with cutoff to return up to the N most probable predictions that are greater than or equal to cutoff. If costs are used, specify bestN with cutoff to return up to the N least costly predictions that are less than or equal to cutoff.

cost_matrix_clause

You can specify cost_matrix_clause as a biasing factor for minimizing the most harmful kinds of misclassifications. cost_matrix_clause behaves as described for "PREDICTION_COST".

Syntax Choice

PREDICTION_SET can score the data by applying 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 prediction_set syntax to score the data with a pre-defined model. Supply the name of a model that performs classification or anomaly detection.

    Use the prediction_set_ordered syntax for a model that requires ordered data, such as an MSET-SPRT model. The prediction_set_ordered syntax requires an order_by_clause clause.

    Restrictions on the prediction_set_ordered syntax are that you cannot use it in the WHERE clause of a query. Also, you cannot use a query_partition_clause or a windowing_clause with the prediction_set_ordered syntax.

    For details about the order_by_clause, see "Analytic Functions" in Oracle Database SQL Language Reference.

  • Analytic Syntax: Use the analytic syntax to score the data without a pre-defined model. The analytic syntax uses mining_analytic_clause, which specifies if the data should be partitioned for multiple model builds. The mining_analytic_clause supports a query_partition_clause and an order_by_clause. (See "analytic_clause::=".)

    • For classification, specify FOR expr, where expr is an expression that identifies a target column that has a character data type.

    • For anomaly detection, specify the keywords OF ANOMALY.

The syntax of the PREDICTION_SET function can use an optional GROUPING hint when scoring a partitioned model. See GROUPING Hint.

mining_attribute_clause

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".)

See Also:

Note:

The following example is excerpted from the Oracle Machine Learning for SQL examples. For more information about the examples, see Appendix A in Oracle Machine Learning for SQL User’s Guide.

Example

This example lists the probability and cost that customers with ID less than 100006 will use an affinity card. This example has a binary target, but such a query is also useful for multiclass classification such as low, medium, and high.

SELECT T.cust_id, S.prediction, S.probability, S.cost
  FROM (SELECT cust_id,
               PREDICTION_SET(dt_sh_clas_sample COST MODEL USING *) pset
          FROM mining_data_apply_v
         WHERE cust_id < 100006) T,
       TABLE(T.pset) S
ORDER BY cust_id, S.prediction;
 
   CUST_ID PREDICTION  PROBABILITY         COST
---------- ---------- ------------ ------------
    100001          0   .966183575   .270531401
    100001          1   .033816425   .966183575
    100002          0   .740384615  2.076923077
    100002          1   .259615385   .740384615
    100003          0   .909090909   .727272727
    100003          1   .090909091   .909090909
    100004          0   .909090909   .727272727
    100004          1   .090909091   .909090909
    100005          0   .272357724  5.821138211
    100005          1   .727642276   .272357724