基于深度模块训练的数据库查询效率预测研究  

Research on Prediction of Database Query Efficiency based on Depth Module Training

作　　者：陈海宇 CHEN Haiyu(School of Public Health,Zhaoqing Medical College,Zhaoqing 526020,China)

机构地区：[1]肇庆医学高等专科学校公共卫生学院,广东肇庆526020

出　　处：《成都工业学院学报》2024年第1期42-46,共5页Journal of Chengdu Technological University

基　　金：广东省教育科学规划课题(2022GXJK650);广东省高等职业院校医药卫生专业教学指导委员会教学改革课题(2022LX030);肇庆市科技创新指导类项目(2023040303003);肇庆医学高等专科学校党建研究课题(2022D27)。

摘　　要：针对现有数据库查询算法效率低、预测精度不足等问题,提出一种基于深度模块的数据库查询效率预测算法。先基于图结构提取待查询数据集的相关特征,采用模块化思维构建深度模块网络模型,通过调整残差块数量和神经网络层数,应对不同规模的数据集并实现模型性能和代价的匹配,明确目标数据点和锚数据点的空间位置关系,通过距离判断提升对目标数据查询预测效率。实验结果表明,所提出算法在应对LUBM5大规模数据集时查询时间控制在50 s之内,同时数据查询的准确率在95%以上。Aiming at the problems of low efficiency and insufficient prediction accuracy of existing database query algorithms,a prediction algorithm of database query efficiency based on depth module was proposed in this paper.First,the relevant features of the data set to be queried were extracted based on the graph structure,and the deep module network model was constructed with modular thinking.By adjusting the number of residual blocks and the number of layers of the neural network,the data sets of different sizes were processed and the model performance and cost were matched,the spatial position relationship between the target data point and the anchor data point was clarified,and the query and prediction efficiency of the target data was improved through distance judgment.The experimental results show that the proposed algorithm can control the query time within 50 s when dealing with LUBM5 large-scale data set,and the accuracy of data query is more than 95%.

关 键 词：深度模块 数据库查询 预测 残差块 锚数据点 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]