Oracle Machine Learning for SQL (OML4SQL) delivers scalable machine learning functionality inside Oracle Database. Build models on large tables in minutes and apply models for “scoring” data in seconds. In Exadata and Autonomous Database, Oracle’s Smart Scan technology pushes scoring processing down to the data storage tier for a significant performance multiplier.
Algorithms are implemented as SQL functions and leverage the strengths of Oracle Database. The SQL data mining functions can mine data tables and views, star schema data including transactional data, aggregations, unstructured data, such as found in the CLOB data type (using Oracle Text to extract tokens) and spatial data. Oracle Machine Learning for SQL functions take full advantage of database parallelism for model build and model apply and honor user data access privileges and security schemes. Predictive models can be included in SQL queries, BI dashboards and embedded in real-time applications.
Oracle Machine Learning also supports a "drag and drop" graphical user interface, Oracle Data Miner, that is integrated with Oracle SQL Developer and is capable of generating SQL scripts from user-created analytics workflows.
In-Database Processing - Data preparation, data transformations, model building, model tuning, and model "scoring" execute as PL/SQL procedures and SQL functions
Machine Learning Algorithms - Users can take advantage of Oracle Machine Learning in-database, parallel algorithms using the SQL language
Familiar SQL - Oracle professionals familiar with SQL can readily use machine learning algorithms as a natural extension to their skill sets
Oracle Data Management - Manage and invoke SQL and PL/SQL scripts in Oracle Database where machine learning objects and functionality are peer to others in Oracle Database - critical for enterprise solutions and applications - involve Oracle security, encryption, parallel execution, SQL analytical functions, tables and views, auditing features, partitioning, text processing, and spatial and graph analytics
Automated Data Preparation - Automatic processing of input data with intelligent defaults or user overrides support handling missing data, treatment of outliers, processing of unstructured text data, binning, and transformations required by each algorithm
Integrated Text Mining - Algorithms accept text columns from tables and views, and then automates term and theme extraction, which are combined with other predictors in building models and scoring data
Partitioned Models - Users can automatically a model comprised of a set of models, where each component is built on a user-specified partition of the data. Scoring occurs at the integrated model level with automatic selection of appropriate component model