| The goal of data mining is to reveal useful and important information from the database to meet the needs of different users.high utility itemset mining algorithm is widely used in commercial activities,considering the internal utility and external utility of itemset at the same time,the goal is to find all itemset with high utility(such as high profit).However,traditional high utility itemset mining still have issues such as follows:(1)The algorithm only contains itemset and their utility information in the mining results,ignoring the number of itemset,which makes the mining information incomplete.People cannot make decisions based on itemset quantity information.And with the explosion of data,users are more inclined to mine more readable useful information from the data.(2)Since the utility value increases with the increase of length,the utility of the long itemset is necessarily greater than the utility of the short itemset,making the algorithm more inclined to mine the long itemset,and it is unfair to use it as a criterion for mining the itemset.In addition,the algorithm search space is larger and less efficient.(3)The algorithm sets a single threshold for the processed dataset,and all items are treated equally,ignoring that items in the database may have different properties or importance.In addition,it is easy to produce rare item problems.That is,the algorithm sets the threshold too high in the mining process,and can only mine high-profit itemset such as {diamond},and it is easy to lose some itemset that users care about.Setting it too low makes it easy to dig out a large number of useless itemset.This paper conducts research on the above problems,and the main work is as follows:(1)In order to obtain the original quantity information from the results,the concept of fuzzy set theory is applied to high utility itemset mining,and each quantity value is assigned a fuzzy value in the interval of [0,1],which represents the membership level belonging to large,medium and small amount.thereby obtaining more readable information.(2)In order to mine the itemset with suitable length and high utility of each item in the data,and balance the number of long and short itemset in the mining results,a High Average Fuzzy Utility Itemset Mining Algorithm(HAFUIM)was proposed.The algorithm comprehensively considers the relationship between fuzzy utility and itemset length,defines the average fuzzy utility,and filters a large number of invalid long itemset by considering the average fuzzy utility of each item in the itemset to determine whether the itemset is a valid itemset.The algorithm designs an average fuzzy list structure to store the necessary utility information and reduce the number of database scans.Four pruning strategies combined with the overestimation upper limit matrix effectively narrowed the search space.Simulation experiments verify the feasibility and efficiency of the proposed algorithm.(3)Considering the characteristics of different items and user needs,a High Average Fuzzy Utility Itemset Mining Algorithm with Multiple Minimum Utility Thresholds(MHAFUIM)is proposed.The algorithm assigns a minimum utility threshold to each different item,which solves the shortcomings of the single threshold algorithm.The threshold-utility list is designed to avoid repeated scans of the database and individual threshold tables.Because the threshold value for each fuzzy itemset in MHAFUIM algorithm is different,the closure attribute does not apply.In order to reduce the search space,a sorting strategy is adopted,and the average fuzzy utility upper limit closure attribute under multiple thresholds is proposed and proved.Finally,through three pruning strategies,many hopeless itemset are pruned in the early stage,and the mining of high-average fuzzy utility itemset with multiple minimum utility thresholds is effectively realized. |
