Selecting and Filtering Association Rules within a Semantic Technique
Mouhir Mohammed, Youssef Balouki, Taoufiq Gadi · International Review on Computers and Software (IRECOS) · 2016
To tackle the problems of thresholding, redundancy, and overlapping from which the Association Rules (AR) suffer, this paper suggests an approach, namely SEMMDPREF, to discover interesting rules by pruning and filtering them. Our approach introduces the notions of dominance between a rule and that of user preference, in addition to the methods and techniques based on semantic significance to efficiently exploit the Ontology Domain (OD). This three-angled approach aims to select the semantically significant, most dominant and preferential rules. Concerning algorithm evaluation, we use a genuine database. The results issued from our experiments show that the OD consideration allows our approach to: Detect and discard rules which are already known and which may be meaningless too. Be able to decrease considerably the size of AR by keeping the ones satisfying the criteria.