基于历史信息的高效近似查询系统  

Efficient approximate query system based on historical information

作　　者：韩雨钢 马廷淮[2,3] 荣欢 HAN Yu-gang;MA Ting-huai;RONG Huan(School of Computer Science,School of Cyber Science and Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;School of Software,Nanjing University of Information Science and Technology,Nanjing 210044,China;School of Computer Engineering,Jiangsu Ocean University,Lianyungang 222005,China)

机构地区：[1]南京信息工程大学计算机学院网络空间安全学院,江苏南京210044 [2]南京信息工程大学软件学院,江苏南京210044 [3]江苏海洋大学计算机工程学院,江苏连云港222005

出　　处：《计算机工程与设计》2025年第2期578-586,共9页Computer Engineering and Design

基　　金：国家重点研发计划基金项目(2021YFE0104400)。

摘　　要：近似查询处理技术是提高数据库聚合查询效率的重要方法,针对海量二维数据提出一种基于历史查询负载的近似查询系统,引入历史查询信息,通过在历史查询空间中进行命中性检测,提高查询区域偏斜等情况时的效率。针对全局查询,通过空间数据划分方法将完整数据集划分为子区域,组织为树状分片索引结构,实现采样和数据摘要方法的结合,提高查询准确性。实验结果表明,当历史查询记录量达到10~4量级时,查询响应时间仅为传统方法的40%。与传统方法相比,该系统平均相对误差降低了63%。随分片数的增加效果有更大提升,当分片数达64时,其平均相对误差仅为传统方法的10%。Approximate query processing techniques play a pivotal role in enhancing the efficiency of aggregate database queries.An approximate query system was proposed for massive high-dimensional data that leveraged historical query workloads.By incorporating historical query information and conducting hit detection within the historical query space,the query efficiency was effectively improved,addressing issues like query region skewness.For global queries,spatial data partitioning techniques were employed to divide the entire dataset into subregions and they were organized into a tree-like shard index structure.This integration of sampling and data summarization methods enhanced the query accuracy.Experimental results demonstrate that when the volume of historical query records reaches an order of magnitude,the query response time is reduced to only 40%of that of tra-ditional methods.Compared to traditional methods,the proposed system achieves an average relative error reduction of 63%.The effectiveness of the system increases with the increase of shards.When the number of shards reaches 64,the average relative error is only 10%of that of the traditional methods.

关 键 词：数据库系统 近似查询处理 空间索引 历史查询 分片索引树 学习型索引 空间填充曲线 

分 类 号：TP311.132[自动化与计算机技术—计算机软件与理论]