Overview of Data Mining - Java Data Mining: Strategy, Standard, and Practice - page 8

provide a summary of dollar sales by day, week, month, quarter, or
year. Multiple dimensions can be used to form a data cube; for exam-
ple, the store where the sale was made and the product that was sold
are each dimensions, as depicted in Figure 1-2. The sale amount is
called a measure, and a cube may have multiple measures. Once the
cube is defined, business and data analysts can “slice and dice” it to
provide different views of the data (e.g., sales by month by geograph-
ical region by product category). Like querying large databases,
OLAP is also deductive in nature. Users formulate questions or orga-
nize data to retrieve answers. The underlying data representation for
OLAP is often in the form of a star or snowflake schema, as shown in
Figure 1-1.
In contrast, data mining supports knowledge discovery of hidden
patterns and insights. It takes an inductive approach to data analysis,
building up results by analyzing each of potentially millions of
records. Data mining allows the answering of such questions as how
much revenue will each store generate for portable DVD players in
the next quarter, or which customers will purchase a portable DVD
player and why. Although OLAP generally supports trend analysis
and forecasting, it may rely on simple moving average or percentage

Item-4

Item-3

Product

Item-2

Dec-31-06

…

Item-1

Date

Jan-1-06

Store-1 Store-2 Store-3

Store

Figure 1-2 Example of an OLAP cube.