通过计算影响权值实现敏感序列模式隐藏  

Hiding Sensitive Sequential Patterns by Computing Impact Weight

作　　者：华蓓[1] 钟诚[1] 黄肇明[2] 杨柳[1] 

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004 [2]广西医科大学第一附属医院教务部,广西南宁530022

出　　处：《小型微型计算机系统》2010年第8期1647-1651,共5页Journal of Chinese Computer Systems

基　　金：广西科学基金项目(桂科基0575014)资助;广西高校优秀人才计划项目(RC2007004)资助;广西高校人才小高地建设创新团队资助计划项目(桂教人[2007]71号)资助;广西大学科研基金项目(XGL090003)资助

摘　　要：基于隐私保护的数据挖掘是近年来数据挖掘领域的研究热点之一.提出一个基于数据清洗的敏感序列模式隐藏算法,它通过计算事务组影响权值,选取对非敏感序列模式影响最小的事务组进行清洗,从而在确保隐藏敏感序列模式的同时,尽量减少对非敏感模式集的影响.对不同稠密度和不同规模数据集的实验结果表明,该算法既实现了对敏感模式的保护又具有较低的误隐藏率,并且当数据集大小变化时,序列数据库相异度没有明显改变.The data mining based on privacy protection has been one of the hot research topics in recent years.This paper presents a hidden sensitive sequential patterns algorithm using data sanitization.It sanitizes the transactions of the sequence that has minimum impact on the non-sensitive pattern by computing the impact weight of the transactions such that it can hide sensitive sequential patterns and minimize the impact on the non-sensitive pattern set at the same time.The experimental results with the different density and different sizes of data sets show that the presented algorithm can protect the sensitive patterns with a lower mistake hidden rate,and the difference between the original and sanitized sequence database has no significant changes with the changes of data set size.

关 键 词：隐私保护 知识隐藏 序列模式挖掘 数据清洗 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]