并行的关联规则增量式更新优化算法研究

摘要

Pruning and Fast Updating algorithm(PFUP) always overloads the processor and decreases executing efficiency. An improved Parallel,Pruning and Fast Updating(PPFUP) algorithm is presented to solve these problems of PFUP. PPFUP algorithm based on shared-nothing workstations,partitions the locally-large itemsets and assigns them to different workstations,and synchronizes the tasks of computing on two databases to reduce the stress of memory. Simulation shows that the execution time of PPFUP algorithm reduces to 11％ compared with PFUP algorithm in the case of the same database and low support threshold.%针对关联规则快速更新算法(PFUP)会带来单处理机运行负荷过重、降低执行效率的问题,提出基于并行计算快速更新算法(PPFUP).在无共享的工作站上,通过划分局部频繁项集并分配到结点,采用并行思想处理原数据库和新增数据库的计算任务,降低单处理机内存压力.实验证明了该算法的有效性.