Mining Association Rules Restricted on Constraint

The aim of this paper is to solve the problem of mining association rule restricted on a given constraint itemset which often changes. At the time of building the system, on given database, we mine first the lattice of closed itemsets. Based on that lattice, whenever the constraint or the minimum support changes, the lattice of all restricted frequent closed itemsets is obtained. The set of all association rules restricted on constraint partitions into disjoint equivalence classes. Each class is represented by a pair of two nested frequent closed itemsets. Then, we just mine independently each rule class. Users can select the rule class that they are interested in. Spending only a little of time, we can mine and figure out the basic rules of that class. They are useful for users because their left-handed sides are minimal and their right-handed sides are maximal. When necessary, the set of all remaining consequence ones together with their confidences can be quickly generated from the basic ones. This consequence set also splits into the different subsets according to different generating operators. Hence, our approach is very efficient and close to user! The theoretical affirmations and experimental results prove that.