一种改进的流数据上的高效用模式挖掘算法  

Improved Algorithm on High Utility Pattern Mining over Data Stream

作　　者：谢志轩 李玉强[1] 

机构地区：[1]武汉理工大学计算机科学与技术学院,武汉430063

出　　处：《小型微型计算机系统》2017年第9期2080-2085,共6页Journal of Chinese Computer Systems

基　　金：中央高校基本科研业务费专项资金项目(2014-IV-106)资助

摘　　要：为了快速地从无限的流数据中挖掘出高效用模式,基于已有算法HUM-UT提出一种流数据上的高效用模式挖掘算法——IHUM-UT(Improved High Utility Mining based on Utility Tree)算法.IHUM-UT算法通过压缩HUM-UT算法的头表大小,使其只包含滑动窗口中关注的数据,减少挖掘时所要遍历的数据量,达到提高时间效率的目的.结合两个数据集,调节最小效用阈值、批大小和窗口大小,对两个算法进行对比实验,实验结果表明,IHUM-UT算法得到的高效用模式集与HUM-UT算法完全一致,在时间效率上有较大提升,这种提升在关注数据量较少、不同数据项个数较多的情况下更为突出.To mine high utility patterns from infinite data stream efficiently,this paper proposes an algorithm on high utility pattern mining over data streams,IHUMUT,after analyzing an existing algorithm called HUM-UT.Through compressing a data structure called head table and reducing the data which need to be traversed during mining,IHUM-UT becomes more efficient than HUM-UT in the matter of running time because its head table only contains the focused items in the sliding window.The paper conducts some contrast experiments on IHUM-UT and HUM-UT with two data sets through adjusting the minimum utility threshold,the batch′s size and the window′s size.The experimental result shows that IHUM-UT,whose high utility itemset is same to HUM-UT′s mining result,has a great improvement on running time especially under some circumstances.For example,the amount of concerned data is small and the number of different item′s type is large.

关 键 词：流数据 滑动窗口 高效用模式挖掘 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]