Skip to content

Machine Learning & AI Features

Last updated: 2026-04-27 Tags: machine learning, AI, OML, AutoML, explain, natural language, AI assistant, augmented analytics, clustering, regression, classification

📖 Full Oracle Documentation: Visualizing Data — Machine Learning · OCI AI Services

Summary

OAC integrates machine learning and AI capabilities at multiple levels: automated insight discovery (Explain), natural language querying (Ask), generative AI (AI Assistant), in-platform model training (OML via Data Flows), and OCI AI Services integration (Vision, Language, Forecasting). These features make analytics accessible to both business users and data scientists.

Augmented Analytics (Business User AI)

Explain

Automatically analyzes a selected measure and generates insight visualizations:

Access: Right-click a column in a Workbook → Explain

Generated Insight Types: | Insight | What It Shows | |---|---| | Basic Facts | Distribution, total, average, nulls | | Key Drivers | Which dimensions have the highest correlation with the measure | | Segments | Clusters of data points with similar behavior | | Anomalies | Outliers and unexpected values |

Each insight is presented as a visualization; users can add to canvas with one click.

Ask (Natural Language Query)

Type plain English questions to generate visualizations:

Access: Workbook → Ask bar (top of canvas)

Examples: 

"Show revenue by region for 2024"
"Top 10 products by profit margin"
"Compare sales this year vs last year"
"What is the trend for customer count?"

OAC interprets the question, selects appropriate columns, and renders a visualization. Users can then adjust or refine.

Auto Insights

OAC proactively surfaces insights when opening a Workbook:

  • Identifies metrics that changed significantly
  • Detects correlations
  • Surfaces anomalies Access: View → Auto Insights panel

AI Assistant (Generative AI)

Powered by OCI Generative AI Services:

  • Describe: Generate a natural language description of a visualization
  • Ask questions: Conversational Q&A against the dataset
  • Suggest: Recommend visualization types for your data
  • Code generation: Generate Logical SQL from a description

Setup: Enable in Service Console → AI & Machine Learning → Configure OCI GenAI connection

Oracle Machine Learning (OML) in Data Flows

Train and apply ML models using a visual pipeline — no coding required.

Supported Algorithm Categories

Numeric Prediction (Regression) | Algorithm | Best For | |---|---| | Linear Regression | Linear relationships | | Random Forest Regression | Non-linear, high dimensionality | | Gradient Boosting Regression | High accuracy, tabular data | | Neural Network Regression | Complex patterns |

Binary Classification | Algorithm | Best For | |---|---| | Logistic Regression | Interpretable, linear boundary | | Random Forest Classifier | Balanced accuracy | | Gradient Boosting Classifier | High accuracy | | SVM (Support Vector Machine) | High-dimensional | | Naïve Bayes | Text, categorical features |

Multi-Class Classification Same algorithms as binary, extended to multiple classes.

Clustering | Algorithm | Best For | |---|---| | K-Means | Known number of clusters | | Hierarchical Clustering | Exploring cluster structure |

Training a Model (Data Flow)

  1. Create a Data Flow
  2. Add Train Numeric Prediction (or other train step)
  3. Configure:
  4. Target column: what you're predicting
  5. Input columns: features
  6. Algorithm: auto or specific
  7. Model name: saved to Models catalog
  8. Run the Data Flow
  9. View model quality metrics (RMSE, accuracy, AUC, etc.)

Applying a Model (Scoring)

  1. Create a Data Flow
  2. Add your data source
  3. Add Apply Model step → select trained model
  4. Output includes prediction column + probability/confidence
  5. Save as Dataset for visualization

OCI AI Services Integration

OAC can call Oracle Cloud AI Services from within Data Flows.

Oracle AI Vision (Image Analysis)

  • Object detection, image classification, custom model
  • Use case: Analyze product images, detect defects in IoT images

Oracle AI Language

  • Sentiment analysis on text columns
  • Key phrase extraction
  • Named entity recognition (NER)
  • Language detection
  • Text classification

Data Flow Step: Add System Connection → select OCI Language Service → select operation

Oracle AI Forecasting

  • Time-series forecasting service
  • Predicts future values of metrics
  • Use case: Revenue forecasting, demand planning, inventory prediction

Oracle AI Document Understanding

  • Extract text from PDFs and images (OCR)
  • Table extraction from scanned documents
  • Key-value pair extraction

Pre-Built Models vs. Custom Models

Scenario Approach
Standard prediction task (regression/classification) Train via Data Flow (OML)
Complex custom model Train in OCI Data Science → register in OAC
Text/image analysis OCI AI Language / Vision service
Time-series forecasting OCI AI Forecasting service

Registering External Models (OCI Data Science)

Models trained in OCI Data Science (Python, AutoML) can be registered in OAC:

  1. Save model to OCI Model Catalog
  2. OAC → Machine Learning → Register → point to OCI Model Catalog entry
  3. Apply registered model in a Data Flow Apply Model step

Model Management

Home → Machine Learning (navigation item): - View all trained models - Model quality metrics - Related Data Flows - Delete or rename models

Best Practices

💡 Tip: Use Explain on your key business metrics to quickly discover which dimensions are driving changes — no ML expertise needed.

💡 Tip: Always evaluate model quality metrics before deploying predictions. For regression: RMSE, R²; for classification: accuracy, AUC, precision/recall.

⚠️ Warning: OML models trained in Data Flows use OAC's in-memory engine. For production-grade models on large data, use OCI Data Science and register the model.

💡 Tip: Apply Model outputs a prediction score. Join this back to your visualization dataset to show predictions alongside actuals.