摘要
近年来,我国互联网规模不断扩大,数据流量呈现出来源分布广、类型复杂多样的特点,给网络性能监测工作带来了困难。论文提出引用核函数空间对传统的主成分分析PCA算法进行改进,通过改善对数据非线性结构的分析能力来提高针对网络异常流量的检测效率。随后采取误报率和检测率结合的ROC评价指标,对该算法进行了仿真实验,证明了该算法的有效性。
In recent years,the scale of the Internet in China has been expanding continuously,and the data flow has the characteristics of wide distribution of sources and complex types,which brings difficulties to the network performance monitoring. In this paper,kernel function space is introduced to improve the traditional PCA algorithm. By enhancing the analysis ability of data nonlinear structure,the detection efficiency for network abnormal traffic is improved. Then the ROC evaluation index combining false alarm rate with detection rate is used to simulate the algorithm,which proves the effectiveness of the algorithm.
作者
金波
JIN Bo(Information Communication Company State Grid Hubei Electric Power Company,Wuhan 430077)
出处
《计算机与数字工程》
2020年第6期1440-1444,1449,共6页
Computer & Digital Engineering
关键词
异常流量
主成分分析
核函数空间
矩阵分解
ROC
abnormal traffic
principal component analysis
kernel function space
matrix decomposition
ROC
