Abstract: We proposed a method addressing the optimization of the association rules pruning via two kinds of “unexpected” rules defined by the chisquare test and interest. We also provided an algorithm on the associat ion rules pruning. Based on the experiments, the method retrieved some effective and useful new rules which are often ignored by other methods. The experiment results also show that the method has a satisfying flexibility and effectiveness.

Key words: data mining, association rules, rules pruning, rules t emplate, interest measure