Nouvelles représentations concises exactes des motifs rares

Abstract: Until a present, the majority of work in data mining were interested in the extraction of the frequent itemsets and the generation of the frequent association rules from these itemsets. Sometimes, the frequent of associations rules can revealed not-interesting in the direction where a frequent behavior is in general a normal behavior in the database. These last years, some work was focused on the exploitation and the extraction of rare itemset and shows them interest. However, the very important size of those itemset was the handicap of algorithms that exploit the rare pattern. In order to relieve this problem, the present report proposes two exact concise representations of the rare itemset, one based on the minimal generators and the other based on the closed itemset. In this context, we introduce two new algorithms called GMRare and MFRare which extract these two exact concise representations.