|
|
Abstract
To make association rule mining practically useful, it is important to
identify in a set of mined association rules the interesting ones.
Subjective approaches to identify interesting rules require that a domain
expert manually works on the mined association rules, which can be tedious
and prone to error. Objective approaches, such as those based on
information theory, avoid the problem by picking the rules that offer the
most information, according to some objective interestingness measure. In
the seminar, I'll talk about some well-adopted interestingness measures in
information theory, and then go on to discuss an algorithm which prunes
redundant association rules using maximum entropy principle.
Read the Presentation
Slides...
Referred Papers
|
