采用布尔映射矩阵的Apriori算法改进研究

针对Apriori算法在数据挖掘过程中需要产生大量的候选集及重复扫描事务数据库等不足,本文基于事务数据库的布尔映射矩阵,提出一种仅需一次扫描数据库的方法。该方法不需要产生候选项集,通过矩阵行交、列交运算及相似度矩阵行交运算,按照项目维度由大到小的反向迭代方法即可发现频繁项集的布尔映射矩阵改进算法(BMM_IA)。研究与实验表明,改进算法节省内存开销、运算速度快,为关联规则挖掘研究与应用提供了新路径。

Research on the Improvement of Apriori Algorithm based on Boolean Mapping Matrix

In view of the shortage of Apriori algorithm in the process of data mining, which needs to produce a large number of candidate sets and repeatedly scan the transaction database, this paper proposes a new method called BMM_ IA based on the Boolean mapping matrix of the transaction database, which only needs to scan the database once and does not need to produce candidate sets. Through matrix row intersection, column intersection and similarity matrix row intersection, the frequent item set can be found by the reverse iteration method of item dimension from large to small. The research and experiment show that the improved algorithm saves memory cost and fast operation speed, and provides a new path for the research and application of association rule mining.