基于常量条件函数依赖的冲突消解算法  

A Conflict Resolution Method with Constant Conditional Function Dependencies

作　　者：冯钦 曹建军 郑奇斌 张磊 FENG Qin;CAO Jian-jun;ZHENG Qi-bin;ZHANG Lei(Command&Control Engineering College,Army Engineering University,Nanjing Jiangsu 210007,China;The 63rd Research Institute,National University of Defense Technology,Nanjing Jiangsu 210007,China)

机构地区：[1]陆军工程大学指挥控制工程学院,江苏南京210007 [2]国防科技大学第六十三研究所,江苏南京210007

出　　处：《计算机仿真》2020年第2期279-283,共5页Computer Simulation

基　　金：国家自然科学基金资助项目(61371196)。

摘　　要：目前绝大部分冲突消解方法都是基于迭代计算数据源可靠度和事实可信度的机制。当数据源较少时,数据源的可靠度难于进行评估,仅凭投票来消解冲突往往会造成较大误差。针对数据源较少时的冲突消解问题,提出基于常量条件函数依赖的冲突消解算法。根据多个数据源之间的冲突,找出冲突匹配对及对应的冲突候选值集合。考虑常量条件函数依赖中具体到部分实例子集的约束关系,将常量条件函数依赖集作为先验知识,通过判断候选值是否符合常量条件函数依赖来选择正确的候选值,避免了错误数据比例较大时直接投票选择产生的误差。通过两个真实数据集上的对比实验验证了上述算法的有效性。At present,most conflict resolution methods are based on the mechanism of iterative calculation of data source reliability and fact credibility.When there are fewer data sources,the reliability of data sources is difficult to assess,and the resolution of conflicts by voting alone often results in large errors.To solve the conflict resolution problem with fewer data sources,a conflict resolution algorithm with constant conditional function dependencies is proposed.According to the conflicts between multiple data sources,the conflict matching pairs and corresponding conflicting candidate value sets were found.Considering the constraint relation of a partial instance subset in the constant conditional function dependence,the constant conditional function dependence set was used as a prior knowledge to select the correct candidate value by judging whether the candidate is in accordance with the constant conditional function dependence,and avoids the error caused by the popular vote selection when the proportion of the error data is large.Finally,the validity of the algorithm was verified by comparative experiments on two real datasets.

关 键 词：数据清洗 数据质量 数据冲突 常量条件函数依赖 冲突消解 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]