International Academic Journal of Science and Engineering

New parallel algorithm to frequent item set mining based on bit matrix in the form of bilateral process

Jaber Hosseinzadeh and Abdoreza Savadi

Abstract: Frequent item set mining is of the most significant subjects of association rules, which are one of the most important branches of data analysis. Frequent item set mining is the solution of this matter and the key point of the research as well. Primary researches include series of algorithms such as Apriori and FP-growth. Parallel algorithms were introduced after series ones in order to reduce work performance time by increasing the number of item sets. The present article proposes a novel parallel algorithm based on bit matrix in the form of bilateral process. The proposed algorithm tries to reduce work performance time and also take the benefits of frequent and non-frequent sets properties related to decreasing the number of investigable sets by means of mixing up-down and bottom-up problem solving approaches. Moreover, it uses a bilateral process to solve problems and to find frequent and infrequent sets. To be just and have authentic results, the same computer system and circumstances were utilized for the experiments and applications. Analyses and evaluations of the experiments results revealed an approximately 20% promotion in performance time in compared with the source algorithm of this study so that the effectiveness of the proposed algorithm is proved

Keywords: Parallel algorithm, Bilateral process, Frequent item set mining, Bit matrix, Association rules

Page: 201-207

Volume 3, Issue 1, 2016