Association Rules Mining Using Multi-objective Coevolutionary Algorithm

Abstract

Association rule mining can be considered as a multi-objective problem, rather than as a single objective one. To enhance the correlation degree and comprehensibility of association rule, two new measures, including statistical correlation and comprehensibility, as objection functions are proposed in this paper. Their calculating formulas and primary characteristics are given. Association rule mining is generally solved by lexicographic order method. On the basis of discussing the weakness of above method, a new coevolutionary algorithm is put forward in this paper to solve multi-objective optimization problem of association rule. Three coevolutionary operators are designed and the mining algorithm is realized in this paper. According to experimentation, the algorithm has been found suitable for association rule mining of large databases.