Oracle has taken a different approach to analytics. The central philosophy can be summarised as: Instead of taking data to the analytics algorithms, take the algorithms to the data. As a result the analytics capability is embedded in the database. Well Oracle did start life as a database company and this is definitely a different way of thinking about the problem of analytics – specifically data mining, text mining and predictive analytics.
Being different is one thing, being better is another. The approach adopted by Oracle does have some very significant benefits. The most important is the fact that predictive models sit in the very place where they can be used – in the database. Any application that calls data from a database can just as easily call a predictive model, or some other analytics task.
At the present time data mining is seen as a fairly esoteric activity. Certainly the analysts who use this technology needs specialized knowledge. But as with all things, what seems exotic sooner or later becomes commonplace, and the use of predictive models will, probably within five years, become a commonplace activity. Clearly Oracle is already well positioned for this by integrating the algorithms and models into the database.
Oracle Advanced Analytics (OAA) provides two primary mechanisms to support analytical activity:
- Oracle Data Mining is SQL based, and the actual modelling environment, Oracle Data Miner, comes as an extension to Oracle SQL Developer. It supports most of the usual data mining algorithms, and support vector machines feature quite strongly.
- Oracle R Enterprise (ORE) extends database functionality with a library of R functions and makes database tables and views available to the R environment as native R objects. Several notable features include parallelized neural networks, the scoring of database tables, support for time series analysis, and the persistence and management of R objects in the database. Open source R packages can also be incorporated into the environment.
Time series are newly supported in ORE 1.3. This facilitates aggregation and moving window analysis of time series data, with date arithmetic and integration with R time series packages such as fts, tseries, xts, zoo, forecast and arima.
Oracle Data Miner
It looks like yet another graphical, drag and drop model building environment – and indeed it is. What makes this tool particularly easy to use is the way that data, models, workflows and connections are handled, and presented to the user. It’s highly productive, and as all analysts know, the organization of data and models is key to model building. This is not a linear process and an organized environment is critical.
Oracle R Connector for Hadoop
The Oracle R Connector for Hadoop provides access to a Hadoop cluster from R. It can be used on the Oracle Big Data Applicance or on non-Oracle Hadoop clusters. This is part of the Oracle Big Data Connectors suite.
Oracle and R
It sounds like a contradiction in terms – an oxymoron. The bespoke with open source. Despite this, Oracle is clearly trying very hard to accommodate R in all its aspects. New releases of Oracle R Enterprise (ORE) focus on greater compatibility with, and exploitation of R’s capabilities. The details can be found at Oracle’s web site.