Oracle Unveils New Data Science Platform

Oracle rolled out a new data science platform for Oracle Cloud, with a service called Infrastructure Data Science at its core. The system is meant to help data scientists collaborate on and deploy machine learning models.

Oracle said Infrastructure Data Science is different from other offerings because it focuses on teams rather than individual users. Its features include shared projects, model catalogs, team security policies and auditability.  

The company’s attention to collaboration and workflows is in line with the features it unveiled for Oracle HCM last September.

Greg Pavlik, senior vice president of product development for Oracle Data and AI Services, said the new offering will support data science projects by automating workflow and adding team support. Its workflow includes:

• AutoML algorithm selection and tuning, which automates the process of running tests against multiple algorithms and hyperparameter configurations, checks results, and confirms model and configuration selection.
• Automated predictive feature selection simplifies engineering by identifying key predictive features from larger datasets.
• Model evaluation generates evaluation metrics and visualizations to help measure a model’s performance against new data. It can also  rank models over time to enable optimal behavior in production.  
• Automated model explanation provides details on the relative weighting of the factors that go into generating a prediction. For example, a fraud detection mode could enable data scientists to explain which factors are the biggest drivers of fraud. That, in turn, allows processes to be modified and safeguards to be implemented. Oracle said this is the first commercial implementation of model-agnostic explanation.

Oracle’s Data Science Platform

Infrastructure Data Science is one of the Oracle Cloud Data Science Platform’s seven services. The others:

• New machine learning capabilities in Oracle Autonomous Database, with new support for Python and automated machine learning.  
• A data catalog that allows users to discover, organize and trace assets on Oracle Cloud. The catalog has a built-in business glossary to simplify data curation.
• Oracle Big Data Service provides Cloudera Hadoop implementation, with relatively simple management. It also offers machine learning for Spark, so organizations can run Spark machine learning in memory with minimal data movement.
• Oracle Cloud SQL, which enables SQL queries on data in HDFS, Hive, Kafka, NoSQL and Object Storage.
• Oracle Cloud Infrastructure Data Flow, a managed Big Data service that allows users to run Apache Spark applications with no infrastructure.  
• Virtual Machines for Data Science, a set of preconfigured GPU environments with common IDEs, notebooks and frameworks.

Separately, CEO Safra Catz told OpenWorld Europe that Oracle has changed the way it operates as well as its technology stack, according to Diginomica. Oracle’s purpose, she said, is now to “serve” customers in the cloud.

The web site said Catz’s speech displayed a “marked difference” from Oracle’s past approach in messaging. “There was no mention of competitors and the whole presentation felt less antagonistic, more conciliatory,” its report explained. “And whilst Catz spoke at length about the capabilities of Oracle’s AI-enabled systems and ‘second generation cloud’, it appeared that the speech delivered was intended to align Oracle’s own understanding of change with what customers are likely experiencing in the digital enterprise.”

Sign up for our newsletter here.

Image: iStock