摘要： 针对传统海量数据精确查询负载过大的问题，引入基于仿真的置信区间自动抽样方法(Bootstrap)对数据库提供支持。通过对部分或采样数据进行查询，将查询简化到基础数据上，在对整个数据集查询一次的时间内，完成对多个样本重复多次的查询，得到数据库查询的置信区间；再进行基础SQL查询，得到符合用户要求的近似结果。实验结果表明，引入Bootstrap方法进行数据查询是有效的。

关键词: 关系数据库, 近似匹配, 置信区间, Bootstrap方法, 结构化查询语言

Abstract: Traditional accurate query of mass data is easy to lead to overload. Aiming at this problem, this paper uses Bootstrap method of the confidence interval based on the simulation to provide database support. The method reduces the range of query data by querying part or sample data instead of all data sets for simplifying basic data of the query. The time of querying all data sets can complete repeated queries for multiple samples, the confidence bound of the database query should be obtained. Employing a basic Structured Query Language(SQL) query gets the approximation results, which correspond to user demand. Experimental result shows that the Bootstrap method is effective for querying data.

Key words: relational database, approximate matching, confidence bound, Bootstrap method, Structured Query Language(SQL)

中图分类号: 

• TP311