Teradata Data Warehouse Embeds BI ToolsBy Chris Preimesberger | Posted 2007-06-21 Email Print
"Open analytic environment" expands data mining application options while using the Teradata database.Data warehousing giant Teradata on June 19 unveiled new storage management software that allows data mining applications from outside vendors to be embedded and run within the company's proprietary enterprise data warehouse.
Teradata Warehouse Miner 5.1 uses an "open analytic approach" to provide access to best-of-breed data mining technologies from the company's set of global partners, Teradata Vice President of Products and Services Randy Lea told eWEEK.
"It's no longer a competitive advantage to just understand the past by looking in the rear-view mirror," Lea said. "Competitive differentiation comes from predicting what is going to happen through good business analytics."
In addition to embedding partners' applications within Teradata Warehouse Miner, Teradata provides optimization for the data mining process by integrating with partners such as KXEN, SAS, SPSS and Fair Isaac. These companies have standardized on the XML-based PMML (Predictive Model Markup Language) to use Teradata's warehouse.
PMML is an open-standard language that allows customers to share analytic models among applications, enabling them to easily port desktop- and server-based analytic models to run within the Teradata warehouse.
Warehouse Miner 5.1, as with previous versions, integrates data mining within a centralized data warehouse, streamlining the data mining process and making it possible to mine massive volumes of data with improved accuracy and speed, Lea said.
"The addition of predictive analytics to data warehousing and BI [business intelligence] is really common sense, and the message is coming louder and louder," Roger Haddad, chairman and CEO of data mining automator KXEN, told eWEEK.
"However, if building a predictive model takes a few weeks of savvy analysts with an associated cost of $50,000 per model, this will only be considered in very few cases. This is where KXEN plays its key role through automationallowing not only analysts, but also business users, to build models on their own."
Using Warehouse Miner 5.1 and KXEN, users build the analytic data set using the Teraminer analysis engine. Then "quick and good models in large numbers using KXEN" are made, Haddad said, before the models are exported to the data warehouse for scoring purposes and managed by Warehouse Miner's Model Manager.
Teradata's business intelligence software does more than simply describe what happened last quarter or last week to better predict what is going to happen in the future, Lea said. The software uncovers relationships and patterns within the data that can be used to predict behavior and events so that businesses can better understand customer behavior, identify unexpected opportunities, and anticipate opportunities and problems before they happen, he said.
Partners' applications are preconfigured and integrated into a customer's open enterprise warehouse environment, Lea said. One example is neural network algorithms provided by StatSoft. Neural networks are a sophisticated statistical tool capable of identifying and evaluating the interrelationship of a large number of variables at one time. They are capable of self-learning behavior, Lea said.
"Although interesting I do not think it [Warehouse Miner 5.1] will significantly increase market share for Teradata or their data mining application," Gartner Group business intelligence analyst Gareth Herschel told eWEEK.
"What it should do is increase the amount of analytical computation done in the data warehouse [compared with a separate data mining environment] in organizations that already have Teradata deployed. It makes the Teradata deployment you already have marginally more productive and useful, but probably not enough to go out and buy Teradata if you were not already planning on doing so," Herschel said.
Herschel also told eWEEK that Warehouse Miner 5.1, which is available now, offers some efficiency improvements for the less common data mining techniques such as neural networks but is "less significant than Teradata's Analytics Data Set Generator and PMML link for the majority of their data mining/data warehousing customers."
Teradata is currently a division of Dayton, Ohio-based NCR and will be spinning off into its own company later this year, Lea said.
Check out eWEEK.com's for the latest news, reviews and analysis on enterprise and small business storage hardware and software.