Oracle Machine Learning

statistical analysis to advanced, predictive, and prescriptive analytics, including machine learning. If a data scientist builds a predictive model that is determined to be useful and valuable, Information Technology (IT) organizations needs to be involved to figure out deployment. It’s at that point that enterprise deployment and application integration issues become a significant challenge.

In this scenario, to make predictions on larger datasets maintained within a data warehouse, predictive models—and all their associated data preparation and transformation steps—must somehow be translated to SQL and recreated inside the database, or translated to Java for execution in an application. This model translation phase introduces tedious, time-consuming and expensive manual coding steps from the original statistical language (SAS, R, and Python) into SQL and Java. DBAs and IT must somehow “productionize” these separate machine learning models within or alongside the database and/or data warehouse for distribution throughout the enterprise. This is where many advanced analytics projects fail. Some vendors will charge for specialized products and options for just the predictive model deployment capability. Add Hadoop, sensor data, tweets, and expanding big data reservoirs, and the entire data-to-actionable-insights process becomes more challenging.

Oracle Machine Learning