|
Table of
Contents
Abstract
Introduction
Technologies
Data Warehouse
Examples
Characteristics
Platforms
Summary
|
|
1.0 Technologies Related to Data Mining
To make data mining feasible, the appropriate data has to be collected and
stored in a data warehouse, and adequate system resources have to be
available
to make the data mining process feasible.
Statistical analysis systems such as SAS and SPSS have been used by
analysts to
detect unusual patterns and explain patterns using statistical models such
as
linear models. Such systems have their place and will continue to be used.
Data
mining techniques will not replace such analyses but, in fact, may spur
more
directed analyses based on the results of data mining.
Ad hoc querying and report generation are commonly used by many businesses
to
provide input to their decision making. Decision support systems (DSS) and
executive information systems (EIS) are often used for this purpose. At a
more
basic level, report generation tools are used. Applications utilizing the
ad
hoc querying capabilities of relational database systems are ubiquitous.
Data
mining helps focus the use of these systems and techniques so that
relevant
information is obtained faster and the analyst's time is employed more
effectively.
Multidimensional spreadsheets and databases are becoming popular for data
analyses that require summary views of the data along multiple dimensions.
As
with the methodologies described above, the analyst has to provide
continuing
guidance for this process to be successful. Data mining technologies
perform
automatic analysis that can help enhance the value of the data exploration
("drill down", "drill up", summarization along various dimensions)
supported by
multidimensional tools.
Neural networks have been applied successfully in a few applications that
involve classification (see Section 3.2). However, they suffer because the
resulting network is viewed as a black box, and no explanation of the
results
is provided. The lack of such explanations inhibits confidence,
acceptance, and
application of results. Also, neural networks suffer from long learning
times,
which become intolerable when large volumes of data need to be processed.
The enabling power of data mining is even more clearly seen with respect
to
another nascent technology, data visualization, which can make it possible
for
the analyst to gain a deeper, intuitive understanding of the data. Data
mining
can enable the analyst to focus attention on important patterns and trends
and
explore those patterns and trends in depth using visualization techniques.
Data
mining and data visualization work especially well together. Data
visualization
by itself runs the danger of being overwhelmed by the sheer volume of the
data
in commercial databases. Data mining may help by suggesting a starting
point
for fruitful exploration and the appropriate metaphor to use for the
visualization.
Section 3.1 and Section 3.2 present examples of the kinds of discovery
that
data mining technology makes possible. Data mining, while complemented by
the
techniques described above, adds significant value beyond the use of the
traditional techniques.
|
|
