Learn about Oracle Machine Learning for Oracle Database and Big Data, on-premises and Oracle Cloud

  • March 29, 2011

How to Deploy an Oracle Data Miner 11gR2 Workflow Model

Charlie Berger
Sr. Dir. Product Management, Machine Learning, AI and Cognitive Analytics

You've been using Oracle Data Miner have explored your data, built and evaluated a few ODM models and now you would like to automate and or deploy your model(s).  There are several options from the simple using the SQL Developer 3.0 with Oracle Data Miner extension GUI to the complex using the Oracle Data Mining APIs.  Here are a few suggestions:

1.  Using the SQL Developer 3.0/Oracle Data Miner GUI:  First, any Oracle Data Miner workflow can be saved and shared with others for the simplist form of deployment. 

Assuming that you have already built an ODM model, you can also right-mouse-click (rmc) on the Apply node and it offers options to deploy the code for that node as SQL scripts, etc.  See screen shot.

Deployment ss1.jpg2.  Data

Right mouse clicking on Data Transform node flows, will allow you to Create Lineage of all the upstream SQL steps which is useful for deploying your analytical methodology.  I'm copying and pasting a bit from the ODM'r Help below.


Save SQL or SQL script for apply, create table, explore data, or model details. You can either copy to the Microsoft Windows Clipboard or save to a file. Deploy provides this choices:
  • SQL to Clipboard
  • SQL to File
  • SQL script to Clipboard
  • SQL script to File

The SQL that is saved consists of SQL generated by the current node and all of its parents that are data providers. The lineage ends when it encounters nodes that represent persisted items such as tables or models. See Executing Scripts for information about how to run the generated SQL.

Executing Scripts

If you run generated scripts using either SQL*Plus or SQL Worksheet and you prompt users for input, this command before the generated SQL:
set define off

You can either execute this new line separately before you execute the generated SQL or you can generate the new line along with the generated SQL. 

2.  Using the Oracle Data Mining APIs:  As the ODM models are 1st class objects of the Database, you don't need to install the ODMr repository on the production machines.  Instead, you can export the model(s) and associated tranforms to another instance and run them there.

See Oracle® Data Mining Application Developer's Guide 11g Release 2 (11.2)
Part Number E12218-05 documentation which further explains Scoring and Deployment

See Also:  Oracle Data Mining Administrator's Guide for information about exporting and importing data mining models which details for example:

  • Using Oracle Data Pump
  • Database Privileges for Export/Import
  • Directory Object for EXPORT_MODEL and IMPORT_MODEL
  • Tables Created By EXPORT_MODEL and IMPORT_MODEL
  • Tablespace for IMPORT_MODEL
  • Importing From PMML
There is some discussion of deploying models in this Oracle Data Mining 11g Release 2 Mining Star Schemas A Telco Churn Case Study white paper. 
We are considering enhancing the Oralce Data Miner deployment capabiltiies as in the next release and are considering a range of possibilities including from the simple idea of adding a Scheduler Setting to the work flow object.  If you have ideas/suggestions, please forward them along (charlie.berger@oracle.com)

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.