A new way to select the valuable association rules

The extraction of knowledge and mining of useful dependencies represents a well established class of tasks in data mining. Most of the algorithms devoted to this task generate a large number of rules suffering from a problem of choosing a threshold. This paper suggests a new approach that helps to discover significant interesting and relevant rules by adopting simultaneously the notion of dominance between rules and user-preference. Our approach neither favors nor excludes any measures. More importantly, specifications of threshold, for users, are easier to deal with away from any risks. In order to evaluate efficiency of the algorithm, we use a real database, and then we compare our results with the ones issued from other algorithms such as Undominated Rules known as “SkyRules”.