Analyze Data Using Python in Workbench Notebooks

Use Python to analyze data when you need statistical processing, cohort analysis, feature generation, model preparation, or other notebook-based methods. Exact packages and Spark access patterns are dependent on your environment.

To analyze data using Python in a notebook:

1. Load the authorized data you need for your analysis in to the Workbench notebook. When loading data, complete the following guidelines:
   1. Use filters, projections, and aggregations early to reduce unnecessary data movement.
   2. Use spark.sql(...) or spark.table(...) catalog table names with SQL cells where configured.
   3. Use /Volumes/<catalog>/<schema>/<volume>/... paths for files exposed through managed or external catalog volumes.
   4. Use /Workspace/... paths for workspace files.
   5. Use approved oci://<bucket>@<namespace>/... paths when direct Object Storage access is configured.
   6. Join approved customer-provided data when it is available through an approved workspace, Master Catalog, Object Storage, ADW, managed table, external table, managed volume, external volume, or other approved configured path.
2. Save your results in the workspace only when they are temporary and do not need to be discovered outside the workspace. Save reusable outputs to approved designations according to your organization's configurations.

See the Data Engineering section of Using Oracle AI Data Platform Workbench for more information.