Algorithmic Research For Mining Unexpected High Utility Itemsets In Dynamic Environments

In the era of big data,the task of mining target information from huge databases is becoming more and more important.In order to meet the diverse needs of users for data,data mining has become more and more significant,and the generation of high utility itemsets mining has correspondingly solved the problem of not taking into account the number and utility of items.Nonetheless,traditional high utility itemsets mining still have issues such as follows:(1)The potential to lose the profits of unexpected itemsets on a specific slice.(2)The lack of anti-monotony and large amount of computation lead to large search space and low mining efficiency.(3)It does not reflect the real-time update characteristics of the database in reality,resulting in the inability to solve the newly inserted data.This paper conducts research on the above problems,and the main work is as follows:(1)In order to mine itemsets with unexpected profits,we introduce the definition of slices and propose the concept of unexpected high utility itemsets.Slice the cube to get a reference dataset,and mine unexpected high utility itemsets on this reference dataset.That is,to find all unexpected high utility itemsets in a particular slice.(2)We propose an algorithm for mining unexpected high utility itemsets(UHUIM).The algorithm is used to mine unexpected high utility itemsets,which brings unexpected profits to the datasets that need to be analyzed regularly.It proposes the data structure of Unexpected High Utility List,which can store the useful information of itemsets more compactly and reuse the memory during the mining process as well as improve the mining efficiency and saving the storage space.It also proposes the UHUI-Prune strategy,which effectively reduces the search space during the mining process.The algorithm outperforms the ULB-Miner and HUI-Miner algorithms in terms of running time,storage space and scalability when evaluated on five real datasets.(3)We also propose an algorithm for mining unexpected high utility itemsets in dynamic environments(DUHUIM).The algorithm delves into the mining of unexpected high utility itemsets in dynamic databases to solve the problem that traditional static datasets cannot adapt to real-time updates of real-life databases.Aiming at the shortcomings of the UHUIM algorithm in staged mining of unexpected high utility itemsets,the expression of one-stage mining of unexpected high utility itemsets is proposed;In addition,we introduce Dynamic Unexpected Lists(DU-list)as the data structure that efficiently store itemset information,at the same time,the refactoring process is designed to solve the problem of inserting new data in real time in an incremental environment;The algorithm applies the DU-Prune strategy to filter the itemset that does not meet the conditions,which greatly improves the running efficiency of the algorithm.The algorithm outperforms the IIHUM and LIHUP algorithms in terms of running time,storage space and scalability when evaluated on three real datasets of different sizes and different densities,and effectively implements the unexpected high utility itemsets mining in the incremental database.