3-5 Jul 2013 Villeneuve d'Ascq (Lille) (France)

By author > Saidi Rabie

Friday 5
Machine Learning

› 11:30 - 12:00 (30min)
Ranking and selecting association rules based on dominance relationship
Slim Bouker, Rabie Saidi, Engelbert Mephu  1@  , Sadok Ben Yahia@
1 : Institut Supérieur d'Informatique, de Modélisation et de leurs Applications  (ISIMA)
Ministère de l'Enseignement Supérieur et de la Recherche Scientifique

The huge number of association rules represents the main hamper that a decision maker faces. In order to bypass this hamper, an ecient selection of rules has to be performed. Since selection is necessarily based on evaluation, many interestingness measures have been proposed. However, the abundance of these measures gave rise to a new problem, namely the heterogeneity of the evaluation results and this created confusion to the decision. In this respect, we propose a novel approach to discover interesting association rules without favoring or excluding any measure by adopting the notion of dominance between association rules. Our approach bypasses the problem of measure heterogeneity and unveils a compromise between their evaluations. Interestingly enough, the proposed approach also avoids another non-trivial problem which is the threshold value specication.


Online user: 1