数据流中基于滑动窗口的闭序列模式挖掘算法  

Efficient algorithm for mining closed sequential pattern based on sliding window in data stream

作　　者：黄钧钧[1] 谢伙生[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350108

出　　处：《计算机工程与设计》2011年第6期2040-2044,2049,共6页Computer Engineering and Design

基　　金：福建省教育厅基金项目(JB07023);福州大学科技发展基金项目(2006-XQ-22)

摘　　要：为了能够有效地利用滑动窗口技术来挖掘数据流当中的频繁闭序列模式,通过构建CST树(closed sequence tree)来保存频繁闭序列模式及其序列之间的关联关系,研究了滑动窗口在流数据上滑动时,窗口内序列属性以及CST树节点相互关系的变化性质,提出了基于滑动窗口的数据流闭序列模式挖掘算法ECSW(efficient closed sequential pattern mining over streamsliding window)。ECSW充分利用CST树内节点的相互关系,减少了对数据库的扫描,并且能够在不借助其他辅助结构的情况下完成CST树节点的更新。比较了ECSW与SeqStream在不同实验参数下的挖掘效果,实验结果表明,在平均闭序列长度不长时,ECSW有着比SeqStream更好的运行效果。To mine the closed sequential patterns in data stream over sliding window effectively, a structure CST （closed sequence tree） is designed to keep closed sequential patterns in sliding window and relationship among sequential patterns. Properties of sequential patterns and relationship among nodes in CST are studied when the sliding window is moved in date stream. Based on these, an algorithm called ECSW （efficient closed sequential pattern mining over stream sliding window） is developed to mine the closed sequential pattern on sliding window in data stream. The relationship among nodes in CST is used effectively to decrease the operation of scanning database. Besides, CST can be updated without any other accessorial structures. To compare ECSW and SeqStream, experiments with different parameters are presented finally, and the results prove that ECSW can make a better work in the mining of closed sequential pattern over stream sliding window when the average length of the closed sequential patterns is not long.

关 键 词：滑动窗口 频繁模式 闭序列模式 数据流 数据挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]