Changes in This Release for Oracle Machine Learning for SQL Concepts

This is a summary of important changes in Oracle Database 23ai for Oracle Machine Learning for SQL Concepts.

• New Features in 23ai
  The following are the new features and enhancements in Oracle Machine Learning for SQL for Oracle Database 23ai.
• Other Changes
  This is a summary of additional changes in Oracle Machine Learning for SQL Concepts for 23ai

New Features in 23ai

The following are the new features and enhancements in Oracle Machine Learning for SQL for Oracle Database 23ai.

New Features

• Integration of ONNX format models

  Oracle Machine Learning for SQL supports ONNX format models with integration of ONNX Runtime. To learn more, see Oracle Machine Learning for SQL User’s Guide.

• New technique: Embedding

  A new technique called Embedding is introduced. To learn more, see Embedding.

Enhancements

• GLM link functions

  Supports additional Generalized Linear Model (GLM) link functions for logistic regression. The additional link functions are: Logit, Probit, Cloglog, and Cauchit. See About Generalized Linear Model

• XGBoost support for constraints and survival analysis

  Oracle Machine Learning supports XGBoost features such as monotonic and interaction constraints, as well as the AFT model for survival analysis. See XGBoost Feature Constraints and XGBoost AFT Model.

• Embeddings through ESA

  Supports embeddings for the Explicit Semantic Analysis (ESA) algorithm. ESA embeddings enables you to utilize ESA models to generate embeddings for any text or other ESA input. This functionality is equivalent to doc2vec (document to vector representation). See About Explicit Semantic Analysis.

• Multiple Time Series and Time Series Regression

  The Multiple Time Series feature of the Exponential Smoothing algorithm enables conveniently constructing Time Series Regression models, which can include multivariate time series inputs and indicator data like holidays and promotion flags. For details, see Multiple Time Series Models and Time Series Regression.

• Boolean Support

  Oracle Machine Learning supports BOOLEAN data type. See Convert Column Data Types, Numericals Categoricals and Unstructured Text, and Target Attribute in Oracle Machine Learning for SQL User’s Guide.

• Anomaly Detection through Expectation Maximization

  OML4SQL supports distribution-based anomaly detection Expectation Maximization (EM) Anomaly. See Probability Density Estimation, Expectation Maximization for Anomaly Detection, and Anomaly Detection Algorithms. For model settings, see DBMS_DATA_MINING - Algorithm Settings: Expectation Maximization.

• Automated Time Series Model Search

  Enables the algorithm to select the best model type automatically when you do not specify EXSM_MODEL setting. This leads to more accurate forecasting. See Automated Time Series Model Search.

• Database Tables Support Up to 4k Columns

  The database tables can now accommodate up to 4,096 columns. This functionality is referred to as Wide Tables. To enable or disable Wide Tables for your Oracle database, you can use the MAX_COLUMNS parameter. See MAX_COLUMNS.

• Model Includes Data Lineage

  OML4SQL records the query string that was run to specify the build data, within the model's metadata. The build_source column in the all/user/dba_mining_models view enables users to know the data query used to produce the model. See ALL_MINING_MODELS.

• Improved Performance of Partitioned Models

  Performance of partitioned models with high number of partitions and dropping individual models within partition model is improved. To know more about partitioned models, see About Partitioned Models.

• Vector Output for Feature Extraction Algorithms

  Oracle Machine Learning supports VECTOR data type and extension of the VECTOR_EMBEDDING operator for Feature Extraction algorithms to significantly improve the efficiency of working with dense, feature-rich data, making vector-based operations more accessible and efficient for users. To learn more, see Feature Extraction and Scoring.

Changes to the Dictionary Settings, PL/SQL Package, and SQL Functions

Dictionary View Settings

A new function called embedding and algorithm ONNX are added in ALL_MINING_MODELS view. To learn more, see Oracle Database Reference. A new attribute type VECTOR is added to support target attributes of Embedding models. To learn more, see ALL_MINING_MODEL_ATTRIBUTES in Oracle Database Reference.

PL/SQL Package Changes

You can find PL/SQL package specific information in Oracle Database PL/SQL Packages and Types Reference guide.

• New procedures to load and drop ONNX format models are introduced: IMPORT_ONNX_MODELS, LOAD_ONNX_MODELS, and DROP_ONNX_MODEL

• GLMS_LINK_FUNCTION: this setting enables the user to specify the link function for building a generalized linear model. See DBMS_DATA_MINING - Algorithm Settings: Generalized Linear Models.

• The following new settings are added for XGBoost support for constraints and survival analysis.

  Note:

  The XGBoost settings are case sensitive.
  • xgboost_interaction_constraints
  • xgboost_decrease_constraints
  • xgboost_increase_constraints
  • objective: survival:aft
  • xgboost_aft_loss_distribution
  • xgboost_aft_loss_distribution_scale
  • xgboost_aft_right_bound_column_name
  See DBMS_DATA_MINING — Algorithm Settings: XGBoost.
• The following settings are added to support Explicit Semantic Analysis embeddings:
  • ESAS_EMBEDDINGS: when enabled, generates embeddings during scoring for feature extraction models.
  • ESAS_EMBEDDING_SIZE: specifies the size of the vectors representing embeddings.

  See DBMS_DATA_MINING — Algorithm Settings: Explicit Semantic Analysis.

• EMCS_OUTLIER_RATE: identifies the frequency of outliers in the training data. See DBMS_DATA_MINING — Algorithm Settings: Expectation Maximization
• New settings for Exponential Smoothing to support Time Series regression models and initial value optimization for model build:
  • EXSM_SERIES_LIST: setting enables you to forecast up to twenty predictor series in addition to the target series.
  • EXSM_INITVL_OPTIMIZE: determines whether initial values are optimized during model build.

  See DBMS_DATA_MINING — Algorithm Settings: Exponential Smoothing.

• KMNS_WINSORIZE: this setting restricts the data in a window size of six standard deviations around the mean. See DBMS_DATA_MINING - Algorithm Settings: k-Means.

• New Global settings:

  • ODMS_BOXCOX: this setting enables the Box-Cox variance-stabilization transformation.

  • ODMS_EXPLOSION_MIN_SUPP: introduced more efficient data driven encoding for high cardinality categorical columns. You can define minimum support required for the categorical values in explosion mapping.

  See DBMS_DATA_MINING - Global Settings.

New SQL Function

Use VECTOR_EMBEDDING to make inference from ONNX embedding models. To learn about the SQL scoring function, see VECTOR_EMBEDDING.

Other Changes

This is a summary of additional changes in Oracle Machine Learning for SQL Concepts for 23ai

Throughout the document, short descriptions are updated and minor edits are made for better readability.