Machine Learning in Oracle Database – What do you want to do?
Oracle Machine Learning enables building AI applications and dashboards, delivering powerful in-database ML algorithms, automatic ML
functionality, and integration with open-source Python and R. OML algorithms support parallel execution for performance and scalability with
improved memory utilization, and support for explanatory prediction details, partitioned models, and automatic mining of text column.
Classification Clustering
Predict target variable containing 2 (binary) or more (multi-class) category values Group or segment cases into hierarchical clusters producing probabilities, rules,
and statistics
Decision Tree
Generates human-interpretable rules, can be used
for segmentation Automated model search; protection against
Expectation Maximization +
overfitting; numeric and multinomial distributions;
Explicit Semantic Analysis Text categorization suitable for large text corpora high quality probability estimates
Scalable implementation of popular XGBoost Produces specified number, k, of clusters; Euclidean
Extreme Gradient Boosting algorithm; supports tree and linear models Hierarchical K-Means + and cosine distance functions; sparsity
Predict binary (0/1, Yes/No) target attributes with optimizations
Logistic Regression (GLM) + attribute coefficients and model statistics; narrow, Discovers natural clusters up to maximum number
wide, sparse data; enables ridge, feature Hierarchical O-Cluster specified; density-based
selection/generation; row diagnostics Load models from other frameworks for use with
Computes conditional probabilities and yields ONNX-format models in-database ONNX Runtime and SQL operators
Naïve Bayes interpretable probabilities; assumes predictor
attribute independence
Time Series
Well-suited to noisy and complex data, supports
Neural Network +
many hidden layers Forecast or predict sequential numeric data using series order column with
either Number or Date/Timestamp types
Tree-based ensemble method that relies on bagging
Random Forest
and feature randomness
Single, double, and triple exponential smoothing for
Solves linear and non-linear problems; multiple regular and irregular series, with and without trend
Support Vector Machine + solvers; sparsity optimizations; supports multi-target Exponential Smoothing and seasonality; multiple methods supported,
classification (a list of targets per row) including Holt-Winters
ONNX-format models Load models from other frameworks for use with Data preparation feature supporting time series
in-database ONNX Runtime and SQL operators Multiple Time Series regression involving multiple predictor time series,
flag variables, and target time series
Regression
Predict numeric target variable Feature Extraction*
Scalable implementation of popular XGBoost Derive new values where all Input variables considered to generate reduced
Extreme Gradient Boosting algorithm; supports tree and linear models set of variables
Predict numeric target attributes with attribute
coefficients and model statistics; narrow, wide, Text categorization with human-readable topic
Generalized Linear Model + sparse data; enables ridge, feature Explicit Semantic Analysis labels derived from corpus; semantic similarity
selection/generation; row diagnostics estimates among documents
Well-suited to noisy and complex data, supports Non-negative Matrix Derives features based on non-negative linear
Neural Network + Factorization combinations for greater feature interpretability
many hidden layers
Selects “best” set of predictors for linear model; Uses SVD to obtain a set of uncorrelated variables
Principal Component
Linear regression w/feature supports forward, backward, both, and alternate that contain the maximum amount of variance from
Analysis +
selection/generation + direction dataset
Solves linear and non-linear problems; multiple Singular Value Narrow data via tall and skinny solvers; wide data
Support Vector Machine + Decomposition + via stochastic solvers
solvers; sparsity optimizations
Load models from other frameworks for use with
ONNX-format models in-database ONNX Runtime and SQL operators Anomaly Detection
Identify cases as normal or anomalous by learning patterns of normal data
Attribute Importance
Identifies anomalies based on probability density,
Supervised and unsupervised ranking of variables to improve model quality Expectation Maximization + ensuring accurate anomaly detection for better data
integrity
Minimum Description Supports a low-rank SVD-based approach for
Length ranking attribute importance as unsupervised Special case of SVM classification that does not use
method One-Class SVM + a target; Solves linear and non-linear problems;
multiple solvers; sparsity optimizations
Random Forest Computes importance of variables relative to target
Process monitoring to detect anomalies with non-
Unsupervised Pairwise KL Supports unsupervised variable ranking and MSET-SPRT linear, non-parametric patterns in IoT sensor data;
Divergence + pairwise dependency estimates “Multivariate State Estimation Technique”
Select most important variables for classification
CUR Decomposition and regression; Association Rules
Market basket analysis using transactional or 2D data representation to
Ranking extract frequently occurring patterns and rules
Supervised prediction probability of one item ranking over other items
Finds frequent itemsets and generates human-
Extreme Gradient Boosting Supports pairwise and list-wise ranking Apriori interpretable rules; computes support, confidence,
lift, and aggregate measures associated with rules
Row Importance Embedding
Unsupervised ranking of rows
Vector embedding generation using pre-built transformers
Supports low-rank SVD-based approach for ranking Load pre-built transformer bundles with pre- and
CUR Decomposition
row importance as unsupervised method ONNX-format transformers post-processing for use with in-database ONNX
Runtime and SQL prediction operators
Survival Analysis
Examines the time elapsed between one or more events
With Oracle Database 23.7+, supports:
Scalable implementation of popular XGBoost * VECTOR_EMBEDDING output
Extreme Gradient Boosting algorithm; supports Cox proportional hazards model +
Vector input type
and the Accelerate Failure Time (AFT) model
© 2025 Oracle Corporation. All rights reserved. Oracle Machine Learning on Oracle Database 23ai