Tuesday, October 13, 2009

SQL Server Analysis Services Data Mining vs Oracle Data Miner

Data Mining Options with Microsoft SQL Server 2008 Analysis Services are more extensive and better integrated for Business Intelligence Projects than Oracle Data Miner. Both Data Mining offerings require use of Enterprise Editions to gain access to the Data Mining Suite. Unfortunately, the Oracle Data Miner Add-In is not available for users of Oracle XE. In the same vein, users of MS SQL Server 2008 Express(any edition) also lack access to SQL Server Analysis Services.

Given the growing BI communities it seems odd that both Oracle and Microsoft would not offer at least some 'lite' version of their Data Mining Suite to accompany their Database 'free' editions.
In SQL Server 2008 Analysis Services Data Mining (http://www.microsoft.com/sqlserver/2008/en/us/data-mining.aspx) Projects are fairly easily built within the BIDS(Business Intelligence Development Studio) environment which is included with SQL Server Standard and Enterprise Editions. The Enterprise edition contains the complete set of parameterization for the Data Mining algorithms. BIDS is essentially a lite version of Visual Studio, however it handles all Analysis Services Projects including Data Warehouse Design, OLAP Cube development, ETL and other database oriented projects in addition to Data Mining quite well. Once Data Mining Projects are built they can also be viewed, browsed and processed like OLAP Cubes in SQL Server Management Studio as well.

SS 2008 Analysis Services Data Mining Projects have a wide range of algorithms available including Decision Trees, Clustering(EM and K-Means), Linear Regression, Logistic Regression, Naive Bayes, Neural Network and more. Each algorithm has a variety of parameters available via Property Sheets which can be used to tailor the research efforts. All algorithms used in for a particular project are easily browsed, altered and compared against each other yielding an impressive Data Mining research and development platform.

Oracle Data Miner (http://www.oracle.com/technology/products/bi/odm/odminer.html) offers a smaller set of Data Mining algorithm options. Oracle Data Miner is downloaded for free separately for installation in the Enterprise Edition of Oracle Database. Oracle DM is a graphical front end to Oracle Data Mining and like OLAP is closely integrated with the Database Engine(except for XE). Decision Trees , Naive Bayes, Association Rules and Clustering algorithms and others are offered. Oracle also offers an Excel add-in that supports the Data Miner, however it is not as well integrated and featured as the newer Office 2007 Data Mining add-in that is integrated with SQL Server 2008.

Overall the SQL Server 2008 Development Platform is much easier to approach, integrated and to work with than the more disparate Data Mining tool set offered by Oracle.

Classic Oracle Metalink Lives?

For the last several months logins to the Online Oracle Support Site - Metalink have been accompanied by notices of 'Classic' HTML based Metalink ending with the new Flash driven version taking over. My response has typically been to "please keep the Classic Metalink version as an option" - both options have been available online lately.

I was pleased to see today in San Francisco Chronicle online (http://www.sfgate.com/cgi-bin/article.cgi?f=/g/a/2009/10/09/urnidgns002570F3005978D80025764A00499E00.DTL) that I was not alone.. Oracle has evidently backed off dumping the Classic Metalink Portal, however it will be crippled somewhat according to the SFGate article referenced above. I had found the new Flash driven site slow and sometimes buggy. If you don't want or can't install Flash then you would be out of luck under the original plan to end Metalink.

OracleWorld 2009 Not OracleWorld of Past..

Registration snafus, Golf Fairway takes over Exhibit Hall, More Conference Sessions Actually Sales Pitches.. These are some of the Headlines that purused through the Halls Monday Morning at OracleWorld 2009. Attendance seems to be down - not suprising given the economy. Many open parking spots late into the morning at our favorite 5th Street Garage.

Back to headlines.. Many bloggers who registered online in advance were able to spend hours building their session schedules online along with other conference preparation tasks only to find out after attempting to check-in at Registration that their application as a Blogger had not been 'approved' - no reason given -".. we may drop you a email later explaining..".... per the Blogger/Press Registration Desk. It should be interesting reading those Blogs.. evidently Oracle not interested in good Blogger relations.

In the West Building Exhibit Hall about 20% of the Hall is taken over by a Golf Fairway.. fun for many - but was that to fill in for unsold Exhibit Hall spaces? Like last year many grumblings about Oracle Speaker led conference sessions thin on content but rich on sales pitches.