基于CUR矩阵分解的网络异常大数据检测算法  

Algorithm of Detecting Network Abnormal Big Data Based on Cur Matrix Decomposition

作　　者：郑美容[1] Zheng Meirong(Fujian Chuanzheng Communications College,Fuzhou City,Fujian Province 350000)

机构地区：[1]福建船政交通职业学院,福建福州350000

出　　处：《黄河科技学院学报》2022年第2期26-30,共5页Journal of Huanghe S&T College

基　　金：福建省教育厅中青年教师教育科研项目(JZ180358)。

摘　　要：大数据通过网络储存与使用信息,若没有相应的安全防护手段,其信息网络内会出现部分异常数据,使用户隐私安全受到威胁,对此提出一种基于CUR矩阵分解的网络异常大数据检测算法,通过二进尺度参数分析数据信号特性,依靠小波模极大值去除数据内干扰噪声,拟定正常区间与观测区间描述数据特征,构建异常大数据检测指标,将大数据转变成二进制表示形式组建邻接矩阵,重构数据获得残差矩阵,以上述两种矩阵和对应参数当作输入,交替更新矩阵,得到数据内残差数值,结合检测指标实现对网络异常大数据的检测。实验证明,所提方法的检测精准度高,在存在白噪声的状况下依然能够成功检测出网络大数据内存在的异常数据,抗干扰性强。Big data stores information and uses it through the network.If there is no corresponding security protection method, some abnormal data will appear in the information network, which will threaten user privacy and security.For this purpose, a network abnormal big data detection algorithm based on CUR matrix decomposition is proposed.The characteristics of the data signal are analyzed through binary scale parameters, the interference noise in the data is removed by the wavelet modulus maximum, the normal interval and the observation interval is drawn up to describe the data characteristics, construct the abnormal big data detection index, and the big data are converted into a binary representation to form adjacency Matrix, then the data are reconstructed to obtain the residual matrix.The above two kinds of matrices and corresponding parameters are taken as input to alternately update the matrix to obtain the residual value in the data, and to realize the detection of abnormal large data in the network combining the detection indicators.Experiments have proved that the proposed method has high detection accuracy, and can still successfully detect abnormal data in the network big data under the condition of white noise, and has strong anti-interference.

关 键 词：CUR矩阵 异常大数据 小波模极大值 数据特征 残差矩阵 

分 类 号：TU92[建筑科学—建筑设计及理论]