|
Real-Time Intelligence Scores Over OLAP
IBM’s DB2 Intelligent Miner Scoring 7.1 extends database
By Douglas Finlay
April 15, 2001 — Tired of the iterative analysis and querying inherent in SQL online analytical processing (OLAP) that ultimately arrives at only one answer per query? IBM claims that its new DB2 Intelligent Miner Scoring (IMS) version 7.1 can provide real-time relational data-mining analyses and scoring based on just one query. In addition to this data-mining feature for developers, DB2 IMS 7.1 follows the Predictive Model Markup Language (PMML) 1.1 from the Data Mining Group to enable the data models constructed to be shared by other data models also using PMML, regardless of the database.
“There is no shortage of analysis tools, but data-mining capabilities such as clustering, classification and neural network analyses go well beyond what OLAP can do,” said Jeff Jones, senior program manager for IBM’s Data Management Solutions Group. He said integration of DB2 IMS 7.1 into the database engine extends the engine to enable real-time data mining at will against any collection of data. “DB2 IMS 7.1 extends the database engine in the same way there are extenders for text, audioand video,” Jones said.
Although he said that data mining and querying run hand-in-hand in OLAP, in that mining is done to validate certain queries to take them to the next level, similarities end there because, with data mining, “developers don’t know in advance what they will be looking at when they construct the query.” He said in iterative analysis all questions are well known in order to draw out a specific response. “Data mining features more complicated algorithms for neural networking, clustering, segmentation and classifications that are ahead of where OLAP is,” Jones continued.
Because of its ability to deduce patterns in data in real time from queries, integrating scoring into the database engine to data mine “eliminates the notion of data mining that requires mainframes, overnight batch runs and tremendous amounts of data,” Jones said.
Dan Vesset, senior analyst at IDC Corp., said the idea of DB2 IMS 7.1 is to bring data mining to developers rather than keeping it in the hands of a few statisticians and Ph.D.s who until now have had the role of interpreting the data. “Just using OLAP and queries is not enough. Developers could embed data mining into customer relationship management systems to look for patterns indiscernible to the eye,” he said.
Jan Mrasek, senior manager for business intelligence solutions at the Bank of Montreal, said the bank currently uses DB2 IMS 7.1 as a discovery process, to learn about customer behaviors and how the bank might determine which kinds of products it can offer customers based on those behaviors. He said developers build data models and then translate them into PMML and pass them onto the DB2 database. “It executes the model scoring in parallel over 12 processors.” He said it automates the scoring process by helping define structures, thus eliminating heavy transformational work.
Available immediately, DB2 IMS 7.1 costs $15,000 per CPU. An Oracle cartridge for the company’s 8i database is similarly available at $15,000.
|
|
Advertisement
| A PMML Primer |
The Predictive Model Markup Language (PMML) is an XML-based language providing a way for companies to define predictive models and share models between compliant vendors’ applications. It provides applications with a vendor-independent method of defining models so that proprietary issues and incompatibilities are removed to enable the exchange of models between applications.
It permits users to develop models within one vendor’s application and use other vendors’ applications to visualize, analyze, evaluate or otherwise use the models. Previously, this was virtually impossible. However, with PMML, the exchange of models between compliant applications now will be seamless.
Because PMML is based on XML, it comes in the form of an XML Document Type Definition. The new language is the creation of the Data Mining Group (www.dmg.org), a vendor consortium whose members include Angoss Software Corp., IBM Corp., Magnify Inc., NCR Corp., Oracle Corp., SPSS Inc. and the University of Chicago’s National Center for Data Mining.
—Douglas Finlay |
|
Back to Top
|