Oracle Data Mining Technology

One of my kids asked me yesterday why I hadn't updated my blog for almost a month, and while its no excuse, the last 30 days have been filled with nonstop learning about Oracle's products, meeting with customers, and just a bit of imagining the possibilities. One of the Oracle technologies I've spent a lot of time thinking about is Oracle Data Mining (ODM). In an odd twist of fate, both ODM and Sun can trace some of their lineage to the 1980's supercomputer company Thinking Machines.

When Thinking Machines went bankrupt in 1994, the hardware assets of the company and many of the employees were acquired by Sun Microsystems. What remained of Thinking Machines reformed as a data mining software company and developed the Darwin data mining toolkit. Then in 1999, the data mining business was purchased by Oracle and eventually became ODM.

ODM provides a broad suite of data mining techniques and algorithms to solve many types of business problems. including clssificaiton, regression, attribute importance, association, and feature extraction. There are of course many different data mining software packages in existence that could, for instance, determine the association between frequency of an employee's new blog entries and their number of days traveling in a month. Most of those tools would require you to extract records from a database, input them into the data mining package, run the analysis, and eventually probably store the results back into the database. Therein lies one of the unique advantages of ODM. Much of the data that large enterprises want to mine already exists in a database, so why not put the data mining algorithms into the database too, then you wouldn't have to move the data in order to mine it. That is exactly what Oracle did about a decade ago with ODM, and its been evolving ever since.

Today, perhaps the ultimate data mining platform is Oracle's Exadata Database Machine. Much has been written about Exadata's smart flash cache, its hybrid columnar compression, and its fully redundant QDR InfiniBand networking which, combined, make Exadata both a great data warehouse and a great OLTP platform. Add ODM, and Exadata becomes a great platform for such data mining applications as anomaly analysis for fraud analysis, clustering analysis for life sciences drug discovery, or association analysis for product bundling or in-store placement analysis.

You won't need a PhD in statistics to use ODM, but I would recommend the book Super Crunchers to get you started on imagining the possibilities.

Comments:

Post a Comment:
Comments are closed for this entry.
About

marchamilton

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today