Information Technology Reference
In-Depth Information
Chapter 12
Association Rules
12.1 Introduction
The purpose of the association rules mining is to find out a hidden associated
relation between different sets of data fields in a database. It is one of the most
widely used and more studied data mining methods. The association rule model
is an important model among knowledge models in data mining. The concept of
the association rule was proposed by Agrawal, Imielinski, Swami(Agrawal,
1993). It is a kind of simple but very practical rule about data relations. The
model of association rule belongs to description pattern and the algorithm which
discovers the association rule belongs to nonsupervised learning.
At the present
the trends of the research on association rules are as follows:
(1) Discover association rules from the concept with single layer to multiple
layers. That is, the mining rule should effect on different layers of database. For
example, in the analysis of sell database of a supermarket, if only the original
fields of table, such as “bread”, “milk” were mined, it will be difficult to
discovery interesting rules. When some abstract concept about original fields,
such as “food” were considered for mining meantime, some new and abstract
rules will be probably found. So the mining on different abstract layers in
database to discovery rules and metarules is a new research direction.
(2) Improve the efficiency of mining algorithm. Usually large quantities of data
were processed in association rule mining and database may be scanned for many
times. So it is important to improve the efficiency of mining algorithm. There
exist three methods: one is to reduce the times of scanning database and it would
430
