The Oracle Advanced Analytics option is enabled by default during installation of Oracle Database Enterprise Edition. After installation, you can use the command-line utility chopt to enable or disable a database option. For instructions, see "Enabling and Disabling Database Options After Installation" in the installation guide for your platform.

Before upgrading, you must drop any data mining models that were created in Java and any mining activities that were created in Oracle Data Miner Classic (the earlier version of Oracle Data Miner).

After you complete the "Pre-Upgrade Steps", all models and mining metadata are fully integrated with the Oracle Database upgrade process ­ whether you are upgrading from 11g or from 10g releases.

Upgraded models continue to work as they did in prior releases. Both upgraded models and new models that you create in the upgraded environment can make use of the new mining functionality introduced in the new release.

To upgrade a database, you can use Database Upgrade Assistant (DBUA) or you can perform a manual upgrade using export/import utilities.

After upgrading the database, check the DBA_MINING_MODELS view in the upgraded database. The newly upgraded mining models must be listed in this view.

After you have verified the upgrade and confirmed that there is no need to downgrade, you must set the initialization parameter COMPATIBLE to 12.1.

Before downgrading the Oracle Database 12c database back to the previous version, ensure that no Singular Value Decomposition models or Expectation Maximization models are present. These algorithms are only available in Oracle Database 12c. Use the DBMS_DATA_MINING.DROP_MODEL routine to drop these models before downgrading. If you do not do this, the database downgrade process terminates.

Issue the following SQL statement in SYS to verify the downgrade:

SQL>SELECT o.name FROM sys.model$ m, sys.obj$ o 
                  WHERE m.obj#=o.obj# AND m.version=2;

Options for exporting and importing mining models are described in the following table.

The examples use the directory object dmdir shown in Example 28-1 and two schemas, dm1 and dm2. Both schemas have data mining privileges. dm1 has two models. dm2 has one model.

SELECT owner, model_name, mining_function, algorithm FROM all_mining_models;
 
OWNER      MODEL_NAME           MINING_FUNCTION      ALGORITHM
---------- -------------------- -------------------- --------------------------
DM1        EM_SH_CLUS_SAMPLE    CLUSTERING           EXPECTATION_MAXIMIZATION
DM1        DT_SH_CLAS_SAMPLE    CLASSIFICATION       DECISION_TREE
DM2        SVD_SH_SAMPLE        FEATURE_EXTRACTION   SINGULAR_VALUE_DECOMP

PMML is an XML-based standard specified by the Data Mining Group (http://www.dmg.org). Applications that are PMML-compliant can deploy PMML-compliant models that were created by any vendor. Oracle Data Mining supports the core features of PMML 3.1 for regression models.

You can import regression models represented in PMML. The models must be of type RegressionModel, either linear regression or binary logistic regression.

A Data Mining user is a database user account that has privileges for performing data mining activities. Example 28-6 shows how to create a database user. Example 28-7 shows how to assign data mining privileges to the user.

A system privilege confers the right to perform a particular action in the database or to perform an action on a type of schema objects. For example, the privileges to create tablespaces and to delete the rows of any table in a database are system privileges.

You can perform specific operations on mining models in other schemas if you have the appropriate system privileges. For example, CREATE ANY MINING MODEL enables you to create models in other schemas. SELECT ANY MINING MODEL enables you to apply models that reside in other schemas. You can add comments to models if you have the COMMENT ANY MINING MODEL privilege.

To grant a system privilege, you must either have been granted the system privilege with the ADMIN OPTION or have been granted the GRANT ANY PRIVILEGE system privilege.

The system privileges listed in the following table control operations on mining models.

An object privilege confers the right to perform a particular action on a specific schema object. For example, the privilege to delete rows from the SH.PRODUCTS table is an example of an object privilege.

You automatically have all object privileges for schema objects in your own schema. You can grant object privilege on objects in your own schema to other users or roles.

The object privileges listed in the following table control operations on specific mining models.

Comments can be used to associate descriptive information with a database object. You can associate a comment with a mining model using a SQL COMMENT statement.

COMMENT ON MINING MODEL schema_name.model_name IS string;

To drop a comment, set it to the empty '' string.

The following statement adds a comment to the model DT_SH_CLAS_SAMPLE in your own schema.

COMMENT ON MINING MODEL dt_sh_clas_sample IS
           'Decision Tree model predicts promotion response';

You can view the comment by querying the catalog view USER_MINING_MODELS.

SELECT model_name, mining_function, algorithm, comments FROM user_mining_models; 

MODEL_NAME        MINING_FUNCTION  ALGORITHM      COMMENTS
----------------- ---------------- -------------- -----------------------------------------------
DT_SH_CLAS_SAMPLE CLASSIFICATION   DECISION_TREE  Decision Tree model predicts promotion response 

To drop this comment from the database, issue the following statement:

COMMENT ON MINING MODEL dt_sh_clas_sample '';

The Oracle Database auditing system is a powerful, highly configurable tool for tracking operations on schema objects in a production environment. The auditing system can be used to track operations on data mining models.

Unified auditing is documented in Oracle Database Security Guide. However, the full unified auditing system is not enabled by default. Instructions for migrating to unified auditing are provided in Oracle Database Upgrade Guide.