基于标签序列的半结构化数据相似度度量  

Structural similarity measurement for semi-structured data based on tag sequences

作　　者：张利军[1] 李战怀[1] 李宁[1] 李霞[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《华中科技大学学报（自然科学版）》2012年第8期77-81,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划资助项目(2009AA1Z134);国家自然科学基金资助项目(60803043;60970070);西北工业大学基础研究基金资助项目(JC20110225;JC201261)

摘　　要：针对基于路径的半结构化数据结构相似度度量方法不能很好地处理路径部分相似以及忽略了元素之间兄弟关系的问题,提出一种基于频繁关联标签序列的结构相似度度量方法,该方法将半结构化数据的结构信息视为标签序列的集合,采用数据挖掘技术中频繁模式和关联项集的概念及算法,从半结构化数据中挖掘频繁关联标签序列并以此作为特征计算其结构相似度.实验结果证明:提出的基于频繁关联标签序列的半结构化数据结构相似度度量方法可以解决基于路径方法的不足,计算的结构相似度更准确、更合理.As existing path-based structural similarity measurement for semi-structured data approaches cannot handled well the case of matching partially between paths and the sibling relationships between elements are ignored,a new structural similarity measure approach based on association tag sequences was proposed.The structure information in semi-structured data were regarded as set of tag sequences,and adapting the concept and algorithm of frequent pattern and association pattern in data mining field,frequent association tag sequence（FATS） can be mined from semi-structured data collection,and then FATS were used as features of semi-structured data to calculate the structural similarity.Experimental results show that FATS approach is more accurate and reasonable than existing path-based approaches.

关 键 词：数据挖掘 数据管理 半结构化数据 结构相似度 频繁关联标签序列 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]