Practical Applications of Data Mining

2.3 attriBute-orienteD rule generalization 35

with TIDs 3 and 4 because they do not contain any itemset in C

3

. The

candidate {1, 3, 4} in C

3

turns out to be large and is the only member in

L

3

. We terminate the algorithm since C

4

is empty after applying Apriori-

gen( ) with L

3

.

2.3 ATTRIBUTE-ORIENTED RULE GENERALIZATION

The Apriori algorithm discussed in Section 2.2 deals mainly with market

basket data in which each record consists of a set of transaction items. The size

of each record varies because each record contains a variable-length list of

items purchased in each transaction. This method may not be appropriate for

knowledge discovery in relational databases where each record is represented

in terms of a fixed number of attributes. ...