Abstract: The existing cyclic association rules have disadvantage to compartmentalize a cycle into several time segments and the base arithmetic disadvantage is low-level efficiency etc. This paper presents CARDSATSV. It chooses the time sequence vector which consists of the support of item to cluster, and uses DB Index to determine the optimal class number of cluster. It brings forward Cyclic FP-tree(CFP-tree) to discover cyclic association rules. CFP-tree handle cycle clipping technology is based on conditional FP-tree to improve efficiency. Experiments show that CARDSATSV can discover more useful cyclic association rules and can improve efficiency, compared with the existing cyclic association rules.

Key words: time series vector, lusty cyclic association rules, difference sequence arithmetic, Cyclic FP-tree (CFP-treee) algorithm, clustering arithmetic based on difference sequence

摘要： 针对目前周期关联规则难以划分时间区域和基础算法效率低等问题，提出一种基于周期关联规则的发现算法(CARDSATSV)。采用由项目支持度组成的时序向量作为时域数据特征点进行聚类，用DB Index准则控制聚类个数以达到最佳的聚类效果。给出CFP-tree算法来发现周期关联规则，利用基于条件FP-tree 的周期性剪裁技术提高算法效率。实验表明，和目前周期关联规则发现算法相比，CARDSATSV可以发现更多有用的周期关联规则，时空效率有一定的提高。

关键词: 时序向量, 强周期关联规则, 差异序列法, 周期FP-tree算法, 差异序列聚类算法

CLC Number: 

• TP391.4