Trusted Digital Transformation Partner

Oracle Analytics Cloud + AI: Machine Learning Pipelines Without Writing Code

Article By : Rostan Team
Jan 24, 2024

Share This:

Oracle Analytics Cloud (OAC) now embeds a full AutoML engine, Augmented Analytics features, and OCI Data Science model deployment — all accessible from the same UI your business analysts already use. This post covers the technical pipeline from raw data to a deployed ML model serving predictions into a live dashboard.

The OAC ML Pipeline

The end-to-end pipeline has four stages: Data Preparation → AutoML Training → Model Evaluation → Embedded Prediction. Each stage is accessible via the OAC UI but is fully automatable via the OAC REST API and Oracle ML Services API for CI/CD integration.

AutoML in OAC: Under the Hood

When you trigger AutoML, OAC runs an Oracle Machine Learning (OML) algorithm tournament on Oracle ADW in the background. It tests 10+ algorithms — including SVM, GLM, Random Forest, XGBoost, and Neural Network — selects the winner by validation AUC, and registers the model in the OML Model Catalog:

-- OML SQL equivalent of what OAC AutoML does
BEGIN
  DBMS_DATA_MINING.CREATE_MODEL2(
    model_name          => 'CHURN_RF_v1',
    mining_function     => DBMS_DATA_MINING.CLASSIFICATION,
    data_query          => 'SELECT * FROM ml_train_view',
    set_list            => model_settings,
    case_id_column_name => 'CUSTOMER_ID',
    target_column_name  => 'CHURNED'
  );
END;

Embedded Predictions in Dashboards

Once trained, the model is available as a Prediction function in OAC dataset columns. You add a calculated column:

PREDICT(CHURN_RF_v1 USING customer_age, tenure_months, monthly_spend)

This calls the OML scoring engine in ADW at query time — predictions appear alongside regular KPIs in the same dashboard, refreshed with each data load.

Augmented Analytics: Explain Feature

OAC's Explain feature runs 6 statistical analyses in parallel (segment analysis, key drivers, anomalies, correlations, quadrant analysis, forecast) and surfaces natural language insights. Technically it invokes Oracle R Extensibility + OML in the ADW background schema and returns ranked insight cards. For a 5M-row dataset, this runs in under 10 seconds on an ADW medium shape.

Connecting OCI Data Science Models to OAC

For teams who want full Python control (PyTorch, Hugging Face), deploy models to OCI Data Science Model Deployment, then register the HTTPS endpoint in OAC as a Register Model connection. OAC handles the REST call, parses the JSON response, and maps output columns to dataset fields — no APEX or custom code needed.