A SAR-based interesting rule mining algorithm

Abstract

Association rule mining is one of the most important fields in data mining. Rules explosion is a problem of concern, as conventional mining algorithms often produce too many rules for decision makers to digest. This paper discusses how to mine interesting rules with the antecedent constraint being positively associated with the consequent. Notions of simple association rules (SAR), interestingness measures and antecedent constraints are incorporated in the process of interesting rules discovery. The entire set of interesting rules can be derived from the simple rules without any information loss, and the proposed SAR-based mining algorithm performs better than conventional methods by reducing the number of candidate rules.