Institute of Information Theories and Applications FOI ITHEA
Interestingness in Association Rules has been a major topic of research in the past decade. The
reason is that the strength of association rules, i.e. its ability to discover ALL patterns given some thresholds
on support and confidence, is also its weakness. Indeed, a typical association rules analysis on real data often
results in hundreds or thousands of patterns creating a data mining problem of the second order. In other
words, it is not straightforward to determine which of those rules are interesting for the end-user. This paper
provides an overview of some existing measures of interestingness and we will comment on their properties.
In general, interestingness measures can be divided into objective and subjective measures. Objective
measures tend to express interestingness by means of statistical or mathematical criteria, whereas subjective
measures of interestingness aim at capturing more practical criteria that should be taken into account, such as
unexpectedness or actionability of rules. This paper only focusses on objective measures of interestingness.