Fast and Parallel Mining of K High Utility Item Set

---A large number of contributions in the literature have been proposed for item set mining, exploring various measures according to the chosen relevance criteria. However, items are actually different in many aspects in a number of real applications, such as retail marketing, network log, etc. The difference between items makes a strong impact on the decision making in these applications. Therefore, traditional ARM cannot meet the demands arising from these applications. By considering the different values of individual items as utilities, parallel mining focuses on identifying the itemsets with high utilities.The parallel mining of high utility itemsets will take very less time than mining with the single system over large number of transactions. The most studied measure is probably the number of frequent item sets processed in import and export business process. While the problem has been widely studied, only few solutions scale. This is particularly the case when i) the data set is massive, calling for large-scale distribution, ii) the length kk of the informative item set to be discovered is high and/or iii) the data are dynamic. In this paper, we address the problem of parallel mining of large informative kk-High Utility item sets (liki) based on joint entropy. We propose advanced FPHIKS (Fast Parallel Highly Informative kk-High Utility item sets) a highly scalable, parallel liki mining algorithm and forward selection algorithm. FPHIKS renders the mining process of large scale databases (up to double or treble terabytes of data) succinct and effective. Its mining process is made up of only two efficient parallel jobs. With FPHIKS, we provide a set of significant optimizations for calculating the joint entropies of liki having different sizes, which drastically reduces the execution time of the mining process.