Distributed Frequent Itemset Mining

Abstract

Association rule mining is an important technique to discover hidden relationships among items in the transaction. The problem is that association rules are generated by first mining of frequent itemsets in distributed datasets does not gain the best and most accuracy rules.The goal of the thesis is to experimentally finding the most frequent itemsets from distributed data sources which is first phase of association rules generation. Firstly, the global frequent itemsetare generated from global dataset. Secondly, the global datasetare divided into three sites, and then generating the local frequent itemsets from each site. A comprehensive search for the best way to combine the local itemset has been conducted. In this search we find that the union of smallest and biggest of itemsets intersected with the middle always gives result which is equivalent to global itemsets.