A Formal Concept Analysis Approach to Discover Association Rules from Data
Mondher Maddouri · Concept Lattices and their Applications · 2005
The Discovery of association rules is a non-supervised task of data mining. Its mostly hard step is to look for the frequent itemsets embedded into large amounts of data. Based on the theory of Formal Concept Analysis, we suggest that the notion of Formal Concept generalizes the notion of itemset, since it takes into account the itemset (as the intent) and the support (as the cardinality of the extent). Accordingly, we propose a new approach to mine interesting item-sets as the optimal concepts covering a binary table (concept coverage). This approach uses a new quality-criteria of a rule: the gain that generalizes the support criteria.