FREQUENT ITEMSET MINING USING DISTRIBUTED FRAMEWORK

Abstract

Data mining is that the extraction of hidden prognostic data from massive databases, that could be a powerful
new technology with nice potential to assist firms likewise as analysis specialise in the foremost necessary data in their
information warehouses. Data processing tools predict future trends and behaviours, permitting businesses to create
proactive, knowledge-driven choices. Frequent Itemset Mining is one among the classical data processing issues in most
of the information mining applications. It needs terribly massive computations and I/O traffic capability. Resources like
single processor’s memory and processor area unit terribly restricted, that degrades the performance of algorithmic
program. During this paper we've planned associate degree algorithmic program which can run on Hadoop – one among
the recent hottest distributed frameworks that chiefly specialise in Mapreduce paradigm. The planned approach takes
into consideration inherent characteristics of the Apriori algorithmic program associated with the frequent itemset
generation and thru a block-based partitioning uses a dynamic work management. The algorithmic program greatly
enhances the performance and achieves fault tolerance compared to the present distributed Apriori based mostly
approaches. Planned algorithmic program is enforced and tested over a multinode cluster.