The Data Miner: SQL Server Data Mining News (February 2005)
the data miner
SQL Server Data
Volume 2, Issue 1
the February 2005 issue of 'The Data Miner' - the official newsletter of SQL Server
Data Mining. This periodic newsletter keeps you up-to-date with the latest
happenings in the SQL Server Data Mining world and gives you useful tips to
help you get the most from SQL Server Data Mining.
In This Issue:
Plugging Into SQL Server Data Mining
SQL Server Magazine Features "Data Mining Reloaded"
Multiple Data Mining Customers Visit The SQL Customer
What's New In The Upcoming SQL 2005 Community
Data Validation Source Code Now Available
Find Quick Answers To Your SQL Data Mining Questions
SQL Server Data Mining Resources
Plugging Into SQL Server 2005 Data Mining
SQL Server 2005 Data Mining offers an extensible platform
that lets third parties add their own data mining algorithms to our server.
By implementing a small set of interfaces, third-party developers can write
plug-in algorithms that look exactly like native Microsoft algorithms to data
mining application developers as well as to end users. Plug-in algorithms can
take advantage of all the server infrastructure, including storage, security,
multi-user training and querying capabilities as well the design, deployment and
management tools included with SQL Server 2005.
The plug-in architecture for data mining in SQL Server 2005
includes a mechanism for third parties to either re-use one of the built-in
Microsoft viewers for their plug-in algorithms or register their own custom
plug-in viewers for visualizing their trained model content.
Deep integration with OLAP and other SQL Server 2005
components like Integration Services and Reporting Services is another advantage
that comes at zero development cost for ISVs.
The SQL Server Data Mining team has put together several
resources to help third party developers succeed with their plug-in algorithm
SQL Server Magazine Features
"Data Mining Reloaded"
"Data Mining Reloaded", an article by SQL Data Mining Research team members
Alexei Bocharov and Jesper Lind that gives you a broad overview of all the cool,
new data mining features in SQL Server 2005.
Mining Customers Visit The SQL Customer Lab
In the past month, the SQL Lab hosted three customers who
visited Microsoft with SQL Server Data Mining as either their primary agenda or
part of it. The product team helped these beta customers understand new features
and to boost performance - in one case by 6000% - by tweaking their queries to
take advantage of new SQL 2005 DMX language features and by making product
changes. Applications being developed by these customers include:
- cross-sell using the Microsoft_Decision_Trees and
- a data mining score card using the
- a plug-in algorithm in collaboration with Microsoft
Research (see the lead article in this newsletter about plugging into SQL
Server 2005 DM)
What's New In The
Upcoming SQL 2005 Community Technology Preview
There will be an updated Community Technology Preview version
of SQL Server 2005 available to beta customers this month. Look for the
following SQL Server Data Mining updates in this drop:
- New DMX syntax for creating mining structures (CREATE
MINING STRUCTURE) and adding models to existing mining structures (ALTER
MINING STRUCTURE ADD MINING MODEL)
- Two new algorithms: Microsoft_Linear_Regression and Microsoft_Logistic_Regression;
these algorithms are specializations of the Microsoft_Decision_Trees and
Microsoft_Neural_Networks algorithms respectively that were determined by
customer feedback to be important enough to call out explicitly.
- Updates to plug-in interfaces; plug-in algorithm
developers should note that they will need to consume the updated
definitions in the dmalgo.h header file and recompile their code
Data Validation Source Code Now Available
Data validation is another non-obvious application of data
mining that forms developers will find a useful addition to their toolset. A
sample SQL Server Data Mining web application has been available in the
Live!Samples section of SQLServerDataMining.com for a while. We have now made
the source code and the associated PMML model available as well in the
Find Quick Answers
To Your SQL Data Mining Questions
SQLServerDataMining.com now provides an easy interface to
Microsoft's SQL Server Data Mining newsgroup that not only allows you to post
your questions without leaving the site but also lets you do a full-text search
on historical archives. Perhaps you won't even need to post that burning
question and wait for a response from the helpful folks on the forum!
You can find previous issues (as well the current
edition) of this newsletter
The SQL Server Data
or comments? E-mail