Research of mining effective weighted association rules

This paper firstly analyzes the current measures of association rules and then proposes a measure of match as the substitution of confidence. Considering that there are often two kinds of situations in transaction database: (1) the importance of different itemsets are different. (2) When new data are added, how to reflect the popular information of the added database while guaranteeing the correlation between the antecedent and consequent of the rule. In view of the above situations, this paper presents two weighted association rule mining methods, and compares them with the method under the traditional framework of support-confidence. The experimental results show that the antecedent and consequent of the rules generated by the proposed methods have higher correlation and the generated rules also highlight the important itemsets as well as the influences brought by the new data. The proposed methods can make the mined association rules have more practical significance.