SPARK BASED DISTRIBUTED FREQUENT ITEMSET MINING TECHNIQUE FOR BIG DATA

In the phase of association rule mining, frequent itemset mining is an important move. Conventional methods to mine frequent itemsets in the big data epoch pose major obstacles as there is insufficient processing capacity and memory space. This work presents a novel frequent itemset mining technique that utilizes both matrix-based pruning method and Spark. The presented Spark based distributed (SBD) frequent itemset mining technique can efficiently reduce the number of campaign itemsets and also boost the performance of iterative computing. The experimental analysis is carried out using typical benchmark datasets to compare the presented SBD algorithm with the previous algorithm such as parallel FP-growth and distributed frequent itemset mining. The numerical results obtained from the experimental analysis indicate that SBD algorithm has better scalability and efficiency. Furthermore, a case study is accomplished to verify the efficacy of the presented SBD frequent itemset mining technique.