深度置信网络大数据目标属性智能提取仿真  

Simulation of Intelligent Extraction of Big Data Target Attribute in Deep Belief Network

作　　者：张倩[1] 李丽[1] ZHANG Qian;LI Li(College of Humanities&Information,Changchun University of Technology,Changchun Jilin 130000,China)

机构地区：[1]长春工业大学人文信息学院,吉林长春130000

出　　处：《计算机仿真》2023年第10期491-495,共5页Computer Simulation

基　　金：2020年度吉林省教育科学“十三五”规划课题(GH20451);吉林省高教学会2019年高教科研项目(JGJX2019D457)。

摘　　要：利用当前方法提取大数据目标属性时忽略了对大数据的主成分分析预处理,导致提取结果覆盖度低、重复度高。为此提出深度置信网络大数据目标属性智能提取方法。利用主成分分析法构建主成分数学模型,计算数据主成分贡献率完成数据降维,进而简化数据,其次在深度置信网络的基础上构建大数据目标属性提取模型。采用DBM模型将数据进行还原和降噪,获取模型的输入,再利用DBN模型完成训练,实现大数据目标属性智能提取。实验结果表明,所提方法的提取结果覆盖度较高,且所提方法下大数据目标属性智能提取的重复度低,实验结果证明了所提方法应用性能更优。Currently,some methods ignore the principal component analysis for big data,resulting in low coverage rate and high repeatability of results.Therefore,an intelligent method of extracting target properties of big data in deep confidence network was proposed in the paper.Firstly,principal component analysis was adopted to construct a mathematical model of principal components for calculating the contribution rate of principal component and reducing the data dimension,thus simplifying the data.Secondly,a model based on the deep confidence network was constructed for extracting target properties of big data.Then,a DBM model was used to restore and denoise the data,and thus to obtain the input of model.Moreover,the DBN model was used to complete the training.Finally,the intelligent extraction was completed.Experimental results show that the proposed method has higher coverage rate in extraction results and lower repeatability of intelligent extraction of big data target properties.Therefore,this method has better application performance.

关 键 词：深度置信网络 主成分分析 目标属性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]