Machine Learning on the Oracle Cloud

A number of machine learning / AI related announcements were made at the recent Oracle OpenWorld Europe 2020 event in London.    I came away thinking that Oracle probably has more capabilities and experience in the ML/AI market than any other vendor......but the range of options could be confusing to new customers wanting to know where to start!

To help understand the options available, here are the three primary services available on the Oracle Cloud that could be used for machine learning and data science projects


Oracle Machine Learning

  • Machine learning within the Oracle Autonomous Database (ADW / ATP)
  • Extensive array of SQL-based machine learning algorithms along with embedded R
  • Zeppelin Notebook user interface
  • Your data never leaves the database, algorithms benefit from the database's parallel processing and autonomous tuning/scalability  

Oracle Data Science Cloud

  • Enables data scientists to build, train and manage ML models on an open source platform
  • Based on Python, Jupyter Lab (notebooks) and open source ML libraries
  • Collaborative features for Data Scientists to catalog, share and deploy models
  • AutoML features are in the pipeline for model evaluation, explanation and feature selection

Oracle Analytics Cloud

  • Users of Oracle Data Visualization (part of OAC) can build and deploy a variety ML models based on classification, regression and clustering algorithms
  • Also comes with a range of augmented features to "explain" features, suggest appropriate data transformations and perform natural language generation (NLG)
  • Zero coding required, everything achieved via the Oracle Data Viz user interface

So with these 3 options available, which is the best option for a customer wishing to experiment with machine learning or start a new ML/AI initiative?&

Here are some thoughts on how the options suit certain personas and/or use case:

Oracle Machine Learning (OML)

  • New or existing Oracle Autonomous Database customers wishing to build and deploy predictive models using an extensive library of database algorithms
  • Ideal for high-security and large-scale applications as the the data never leaves the Autonomous Database
  • Enables customers to perform machine learning with all the benefits of the Autonomous Database (security, scalability, performance, tuning etc)
  • Caters for all data scientists but is a useful starting point for developers who are comfortable using either SQL, R or Python (coming soon) and would like to get into data science

Oracle Data Science Cloud

  • For any new or existing Oracle customer wishing to kick-start a data science initiative
  • Supports the use of data from a wide variety of data sources in addition to the Autonomous Database
  • Processing can be down off-line (i.e. away from the source database) but still delivers high levels of performance and scalability
  • Enables the use of open source tools and libraries (NOTE: full integration with OML is on the roadmap)
  • Ideal for developers/data scientists who are comfortable with writing code for Python and Spark platforms

Oracle Analytics Cloud (OAC)

  • Ideal for citizen data scientists wishing to utilise machine learning without having to write code and/or provision data science environments
  • Users can run their ML algorithms within the Oracle Analytics Cloud or even on their desktop using Data Viz Desktop (DVD)
  • Supports data coming from a wide variety of data sources (in addition to the Autonomous Database)
  • The graphical UI makes it easy for users to combine the results of their ML "data flows" with other data sets and data flows
  • Useful for rapid prototyping on sample sets of data prior to initiating a full-blown data science project

Naturally the above personas can be challenged and there is certainly a good deal of overlap between the 3 options.    However what I would say is that the Oracle Machine Learning and Data Science Cloud options are more applicable for experienced data scientists who intend to build complex models on large data sets or where multiple models need to be evaluated, assessed and tuned.   OAC however is definitely the best for visualising the results of your predictive models and sharing via storyboards!

Please get in touch if you would like to know more! call us on 01246389000 or email us on enquires@peakindicators.com


Subscribe to our Newsletter

If you enjoyed this article why not get great insights straight to your inbox

Leave a comment