Using appropriate number of computing nodes for parallel mining of frequent patterns

Frequent pattern mining has attracted a lot of attention in past twenty years because of its wide applications like commercial promotion, web search engines, and so forth. However, the execution performance suffers from the rapid growth on the database. Many of the past studies tried to use distributed computing technology to speed up the mining process, but few of them discussed how the appropriate number of computing nodes are located. In this paper, we will propose a novel method named FLB-Mining that is able to effectively determine appropriate number of computing nodes for mining the frequent itemsets. Through empirical evaluations, the proposed method is shown to deliver excellent performance in terms of execution time.