Re: DM: RE: Data Forms for Mining

• Organization: Oracle Corproation

Greg,

Oracle Darwin requires that the data be in a single table of N rows and M columns. The good news is that both N and M can be very large (limited in Darwin only by the amount of disk space and memory). Fortunately, there exists many good tools for performing database table joins, etc. to get the data in a format ready to be mined.

Once the data is in a N x M table, Darwin (an many other data mining applications) provides a rich set of data transformation functions to transform the data to try to bring the information closer to the surface of the data. Also, we support feature selection functions (we call it the "key fields wizard") to identify the importance of each variable as a predictor and to winnow down the candidate variables used in subsequent modeling.

Naturally, as a database vendor Oracle will strive to provide functionality that mitigates this problem and hopefully enhances your ability to deal with common data scenarios.

Greg wrote:

 > Thanks for the information about Darwin.   I will look at that product.   I
 > am a
 > little (lot?) confussed.   Do these products require all the data to be in a
 > single file or can they handle a set of related files?   The answers so far
 > indicate a single file scan.   That would require a design decisioin of 
basic
 > lowest parameter.   All of the records designed around the Customer and an
 > instance with that Customer.  Am I reading this correctly?    When 
designing a
 > database to be used with one of the DM tools, are there basic/general 
processes
 > that go into the design?
 >
 > Greg

--
Charles Berger

Director of Marketing, Data Mining
Oracle Corporation
200 Fifth Street
Waltham, MA 02451
781.684.5841 Office
781.684.5895 FAX
External URL: http://www.oracle.com/datawarehouse/products/datamining/
Internal URL: http://datawarehouse.us.oracle.com/datamining/

• Prev by Date: Re:DM: Anyone doing applied Ontology Discovery via DM?
• Next by Date: RE: DM: RE: Data Forms for Mining
• Prev by thread: RE: DM: RE: Data Forms for Mining
• Next by thread: RE: DM: RE: Data Forms for Mining
• Index(es):
  • Date
  • Thread