2.1 About Oracle Machine Learning Models

Machine learning models are database schema objects that perform machine learning techniques.

As with all schema objects, access to machine learning models is controlled by database privileges. Models can be exported and imported. They support comments and they can be tracked in the Oracle Database auditing system.

Machine learning models are created by the CREATE_MODEL2 or the CREATE_MODEL procedures in the DBMS_DATA_MINING PL/SQL package. Models are created for a specific machine learning technique, and they use a specific algorithm to perform that function. Machine learning technique is a term that refers to a class of machine learning problems to be solved. Examples of machine learning techniques are: regression, classification, attribute importance, clustering, anomaly detection, and feature selection. OML4SQL supports one or more algorithms for each machine learning technique.

Along with the machine learning technique, in the CREATE_MODEL2 procedure, you can specify an algorithm and other characteristics of a model. In CREATE_MODEL procedure you can specify a settings table to specify an algorithm and other characteristics of a model. Some settings are general, some are specific to a machine learning technique, and some are specific to an algorithm.

Note:

Most types of machine learning models can be used to score data. However, it is possible to score data without applying a model. Dynamic scoring and predictive analytics return scoring results without a user-supplied model. They create and apply transient models that are not visible to you.