基于列表结构的加权可擦除项集挖掘算法  

A weighted erasable itemset mining algorithm based on list structure

作　　者：文凯[1,2,3] 许萌萌 张许红 WEN Kai;XU Meng-meng;ZHANG Xu-hong(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065;Research Center of New Telecommunication Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065;Chongqing Information Technology Designing Co.Ltd.,Chongqing 401121,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆邮电大学通信新技术应用研究中心,重庆400065 [3]重庆信科设计有限公司,重庆401121

出　　处：《计算机工程与科学》2021年第9期1676-1683,共8页Computer Engineering & Science

摘　　要：可擦除项集挖掘是从大规模产品数据库中挖掘出低利润项集,以解决厂商财务危机的方法。传统挖掘方法只处理静态产品数据库,在提取可擦除项集时忽略项本身的权值。为解决现有可擦除项集挖掘算法考虑条件单一、效率低下的问题,提出一种有效的在增量数据集上挖掘加权可擦除项集的算法WELI。该算法综合考虑了数据不断积累和项具有不同重要性的因素,采用简洁的列表结构减少内存消耗,利用权重条件进行项集修剪,并结合包含索引和差集思想简化增益的计算过程,以实现高效的增量挖掘操作。实验表明:就运行时间和内存消耗而言,该算法在稠密数据集和稀疏数据集上均具有良好的实验效果,就可伸缩性而言,该算法也优于以往算法。Erasable itemset mining is an approach of mining low profit itemset from large-scale pro-duct databases in order to solve the financial crisis of manufactures.Traditional erasable itemset mining methods deal with static product databases only,and ignore the weight of item itself when they extract the erasable itemset.To address the problem of single condition and the inefficiency of existing erasable itemset mining algorithms,an effective algorithm WELI is proposed to mine erasable itemset in an incremental database with weighted condition.The proposed algorithm comprehensively considers the factors of data accumulation and different importance of items.The concise list structure is applied to reduce the memory consumption.Besides,the proposed algorithm prunes the invalid itemset with the weight conditions.What s more,it can simplify the process of gain calculation by combining subsume index and difference set.Therefore,it can achieve incremental mining operations efficiently.Experiments show that,in terms of running time and memory consumption,the algorithm has good experimental results on both dense and sparse data sets.In terms of scalability,the algorithm is also superior to previous algorithms.

关 键 词：数据挖掘 可擦除项集 增量挖掘 权重条件 包含索引 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]