|
The aim of data mining is the discovery of patterns within data
stored in databases. Mining for association rules is a data mining
method that lends itself to formulating conditional statements such
as "if customers buy product A then they also buy product B and
C with a probability of 90 percent." We consider different
extended concepts of basic association rules. One of these concepts,
quantitative association rules, is discussed in detail. Quantitative
association rules allow statements like "20 percent of
customers who buy at least three units of product A also buy between
five and ten units of B and two units of C." We show that the
quantitative nature of data can be hidden from the algorithm that
mines for association rules. Thus, any standard algorithm that
solves the basic problem can be used to cope with quantitative
association rules. The presentation of our approach includes the
design of the database, algorithms and data structures, as well as
experiments with a prototype implementation.
|