Oracle Machine Learning - no license required

In case anyone missed it, on Dec 5, Oracle announced that customers can now use Oracle Machine Learning (OML) features of the Oracle Database without purchasing additional licenses (previously these features required the purchase of a database option license). 

In this post, I provide a basic introduction to what OML is, why you might like to use it, and where to find more information.  

Many organisations are turning to Machine Learning to extract more value out of ever-increasing amounts of data, and tools like R and Python have never been more popular. These tools have made ML techniques accessible to a broader range of users, but one challenge often encountered is the performance when processing large data sets.  

For example, consider the scenario where a data analyst has a large data set residing in a database and wants to analyse that data to predict customer churn.  

One way to approach this would be to implement a classification model in R. The way this would work is that firstly the database is queried to obtain a sub-set of data so that the dataset is reduced to just that required for the classification model. This resultant dataset is then transferred from the database, across the network, to the client machine. The R client then processes the dataset in memory to run the classification model. Performance issues can arise with the network transfer, and for large datasets, the client system may also struggle with processing.  

With OML (Oracle Machine Learning), the basic process is the same, but the underlying mechanics of how it is executed is different. OML enables customers to run Machine Learning algorithms on data resident in an Oracle database (using SQL, R or Python). In this scenario, the R client is still used but the application of the classification model is performed within the database. This can help in the following ways:

• by moving the algorithm to the data, instead of the other way around, network latency is avoided. 
• for large datasets, the client system may not have sufficient memory to perform the analysis. 
• the underlying Enterprise-class features (performance, scalability, reliability) of the database can be leveraged, which could include:

  - parallel and distributed processing features of the database

  - the Exadata database machine and its plethora of performance features

  - the Oracle In-Memory database option (licensed option)

  - combine analysis with geospatial and graph data (the Oracle Spatial and Graph option can also now be used without purchasing additional licenses)

The following slide shows the range of algorithms available:

Give it a go and see how it works for you. 

Where can I find more info?

The Dec 5 announcement is here: https://blogs.oracle.com/database/machine-learning,-spatial-and-graph-no-license-required-v2

Oracle Machine Learning Overview & Roadmap: https://www.oracle.com/a/tech/docs/oracle-machine-learning-overview-and-roadmap.pdf

Spatial & Graph overview https://blogs.oracle.com/oraclespatial/graph-database-and-analytics-for-everyone

And, as always, feel free to reach out to your local Oracle team as well - we would be happy to run you through the technologies in more detail.