Association Rules Mining Based On Concept Lattice

Knowledge discovery in database is more important area in AI researching now and mining association rules is the widest application in data mining area. The traditional Aprior algorithm can generate all association rules, but the number of association rules is usually very large and redundant because a number of rules can be generated by other rules. The concept lattice that is proposed by Wille et al. in 1982 is an efficent tool of mining association rules. A node in it is called a concept. The extension of concept is a set of object belonged to the concept and the intension of concept is a set of attribute shared by all object. The concept lattice describes the relationship of generalization and specialization among concepts and the Hasse Graph realizes visulization to data .As a kind of formal tool used data analyzing and knowledge processing the concept lattice has already been used in information indexing, data mining, software engineering et al. area.The concept lattice and Hasse Graph expresses the relationship of generalization and specialization among concepts of concept lattice is very suitable for extracting rules. Godin et al. proposed the algorithm of extracting rules based on concept lattice but the number of rules is usually large. Zaki proposed the algorithm that uses the minimal generators of closed items to generate non-redundant association rules, but it has the availablity of losing information.Our algorithm doesn' t extract all rules, but extracts a subset of all rules that is called rule-generating set. We can use it to get all rules. The number of the rule-generating set is smaller than all rules, so the efficency of mining is increased. Our algorithm can' t get each rule' s support and confidence but else algoirthm can . It only gets all rules those support and confidence is larger than threshold that is given...