向量内积策略的多支持度正负关联规则挖掘

Study on mining positive and negative association rules based on vector inner product

在研究负关联规则相关特性的基础上,将向量内积引入到该领域,提出了一种基于向量内积的多最小支持度正负关联规则挖掘算法。考虑到事务数据库中各项集分布不均而导致的单一最小支持度难以设定的问题,采用了多最小支持度策略,设计了一种能同时挖掘出频繁与非频繁项集,以及从这些项集中挖掘出正负关联规则的算法。实验结果表明,该算法仅需扫描一次数据库,且具有动态剪枝,不保留中间候选项和节省大量内存等优点,对事务数据库中负关联规则的挖掘具有重要意义。

Studying on the characteristic of negative association rules,this paper introduces vector inner product to this field, and puts forward a new algorithm to mining positive and negative association rules with multiple minimum supports based on vector inner product.Considering the inhomogeneous distribution of each itemset in transaction database,which may lead to the single minimum support is difficult to be set,it designs an algorithm that can mine frequent and infrequent itemsets,and mine positive and negative association rules from these itemsets with multiple minimum supports.Experimental results show that this method not only scans the database only once,but also has virtues such as pruning dynamically,without saving mid items,and saving lots of memories,which is important to the negative association rule mining in transaction database.