Abstract:

Traditional association rule mining algorithms usually generate association rules that satisfy the predefined minimum support and confidence threshold. However, with the rapid expansion of the data scale, the set of association rules obtained from traditional association rules contains massive redundant rules that cannot reflect the specific interests of users. In this paper, we combine the subjective interestingness optimization algorithm with the objective redundant association rule deletion algorithm in an effort to complement each of their shortcomings and improve optimization ability. We also present an algorithm for non-redundant association rules based on user interest orientation. First, an algorithm is proposed to remove redundant association rules based on the first-order predicate formula. Then, the non-redundant association rule set is used as the original data set for classification. By using the attributes or rules of the user's interest as the guiding object, the template idea is added as the carrier of the user's meaning. Based on the user’s prior knowledge, the template is divided into three types: GI, RPC, and PK. Based on this method, the association rules are obtained using the method of interest calculation. The experimental results show that the algorithm has been improved effectively.