基于cusum算法的电商直播信号异常波动特征建模  

Modeling of abnormal fluctuation characteristics of e-commerce live broadcast signals based on cusum algorithm

作　　者：廖吟秋 王亚春 LIAO Yinqiu;WANG Chunya(China Electronics Technology Company International Trading Co.,Ltd,Beijing 100120,China;Nanjing Customs District Industrial Products Inspection Center,Nanjing 215019,China;Nanjing Normal University,Nanjing 210023,China)

机构地区：[1]中电科技国际贸易有限公司,北京100120 [2]南京海关工业产品检测中心,南京215019 [3]南京师范大学,南京210023

出　　处：《自动化与仪器仪表》2023年第6期54-57,62,共5页Automation & Instrumentation

摘　　要：为进一步提高直播的信号异常检测性能,提出一种基于CUSUM算法的电商直播信号异常波动检测方法。在分析CUSUM算法基本原理基础上,对算法改进优化,然后将算法应用到电商直播信号的异常检测中,并构建异常诊断模型,以解决传统基于CUSUM算法检测准确率低的问题。结果表明,与改进前的CUSUM算法相比,使用基于改进CUSUM算法的检测方法进行增量式以及脉冲式网络攻击时大大提高了检测率,平均检测率能够达到99%以上;与改进前的检测方法相比,使用基于改进CUSUM算法的检测方法进行Flash Crowd现象的检测时,误报率降低了34%。以上结果证明,使用本研究提出的基于改进CUSUM算法的检测方法进行电商直播中的网络异常波动检测时能够保持较高的检测率以及较低的误报率,具有一定的实际应用价值。In order to further improve the network quality of online live broadcasting,based on CUSUM algorithm,the network abnormal fluctuations in e-commerce live broadcasting are studied,and corresponding detection methods are proposed.Among them,CUSUM algorithm is used as the basic detection method of network abnormal signal,and the algorithm is improved and optimized,combined with the improved detection form,to solve the problem of low accuracy of traditional detection algorithm based on CUSUM algorithm.The experimental results show that compared with the improved CUSUM algorithm,the detection rate of incremental and impulsive network attacks using the detection method based on the improved CUSUM algorithm is greatly improved,and the average detection rate can reach above;Compared with the detection method before the improvement,the false alarm rate is reduced by 34%when using the detection method based on the improved CUSUM algorithm to detect the Flash Crowd phenomenon.The above results prove that the detection method based on the improved CUSUM algorithm proposed in this study can maintain a high detection rate and a low false alarm rate when detecting network abnormal fluctuations in e-commerce live broadcast,which has certain practical appliappli-cation value.

关 键 词：网络直播 网络波动 CUSUM算法 网络攻击 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]