A New Dynamic Distributed Algorithm for Frequent Itemsets Mining

Abstract

for association rules between items in large transactional databases is a central problem in the field of knowledge discovery. It has crucial applications in decision support and marketing strategy. Centralized and Distributed Association Rules Mining (DARM) include two phases of frequent itemset extraction and strong rule generation. The most important part of ARM is Frequent Itemsets Mining (FIM)and because of its importance in recent years, there have been many algorithms implemented for it. In this paper, we have focused on distributed Apriori-Like frequent itemsets mining and proposed a distributed algorithm, called New Dynamic Distributed Frequent Itemsets Mining (NDD-FIM), for geographically distributed data sets. NDD-FIM has a merger site to reduce communication overhead and eliminates size of dataset partitions dynamically. The experimental results show that our algorithm generates support counts of candidate itemsets quickerthan other DARM algorithms and reduces the size of average transactions, datasets, and messageexchanges.