加权关联规则算法的并行化研究

Parallel Research of the Weighted Association Rule Algorithm

加权关联规则MINWAL(O)算法有效解决了数据库中各项目的重要程度不同的问题,但在生成加权频繁项集需要多次扫描数据库,遇到大规模数据算法效率很低问题。该文提出一种改进的MINWAL(O)算法,将事务数据库扫描后转化成布尔矩阵,然后对布尔矩阵进行分块,再由多个节点并行计算,并使用多叉树结构存储局部加权频繁项集,最后汇总得出加权频繁项集。该算法与MINWAL(O)相比,减少了数据库扫描次数,提高了算法时间效率。

MINWAL(O)algorithm is an effective solution to the purpose of the database importance of different items,but requires multiple scanning databases.It is ineffective for large databases.Therefore,this paper presents an improved MINWAL(O)algorithm after the transaction database scanning transformed into a Boolean matrix,and Boolean matrix is divided into blocks,and then by a plurality of nodes in parallel computing,and using a multi-fork-tree structure to store locally weighted frequent item set.The final summary results from weighted frequent item sets.The algorithm reduces scan to the database and improves time efficiency of the algorithm.