Overview

As is apparent in the discussion and illustrations throughout Part I, data mining is not married to any particular information technology. The precious information that is the goal of data mining to discover may be apparent in a minute data source, as illustrated by the "case of the collaborating college kids" in Chapter 3. There, a set of 100 rows by 7 columns successfully identified the population of interest. At the opposite end of the spectrum are data mining applications based on massive and sometimes very complex data stores involving millions of records run on very powerful computing platforms. Two examples of this latter type of application are included in Part IV: Fingerhut's Mail Stream Optimization identifies which of its 120 independent catalogs will be sent to more than 6 million customers (Chapter 16), and United Airline's Orion system for passenger-demand forecasting that processes 350,000 logical flight paths for roughly 120 million possible combinations (Chapter 21).

If data mining applications lack specific IT requirements, why then include a chapter on information technology in a book such as this? The answer is to clarify and hopefully simplify the reader's expectation of the IT infrastructure requirements that data mining entails. Also, to provide confidence that some initial data mining initiatives may be carried out without major investment in massively parallel computers and data silos, or expensive special-purpose software tools, as many vendors would prefer for you to believe.

As in any information system, specific technical requirements depend on the nature (size,...