摘要
为了获得更为精准的网络数据流异常检测结果,提升检测效率,减小多因素影响造成的检测误差偏大问题,引入深度长短记忆网络对数据流异常检测参量进行结构优化,优化可分为网络数据流异常特征属性检测量定义、异常数据属性特征提取、深度长短记忆网络下的异常数据特征预处理、建立数据异常检测模型四个部分。通过从异常特征属性定义入手,全方位量化、构建全新深度长短记忆网络下的数据流异常检测模型,以此获得更为精准的检测结果。通过数据对比证实,基于深度长短记忆网络的网络数据流异常检测方法针对一般情况下的异常数据检测灵敏度较高,检测误差低,检测结果可信度和准确率高,检测耗时短,实际应用效果好。
In order to obtain more accurate network data flow abnormal detection results,improve detection efficiency,reduce the detection error of multiple factors,this paper introduces the depth of length memory network to optimize the data flow abnormal detection parameter structure,optimization can be divided into network data flow abnormal attribute detection quantity definition,abnormal data attribute feature extraction,depth under the length of the memory network abnormal data feature preprocessing,establishment of data abnormal detection model of four parts.By starting from the definition of abnormal feature attributes,the data flow anomaly detection model in a new deep memory network is constructed,so as to obtain more accurate detection results.Through data comparison,the network data flow anomaly detection method based on deep length memory network has high sensitivity,low detection error,high reliability and accuracy of detection results,short detection time,and good practical application effect.
作者
周永吉
李阳
黄博
ZHOU Yong-ji;LI Yang;HUANG Bo(Heilongjiang Provincial Meteorological Data Center,Harbin 150030 China;Heihe Meteorological Bureau,Heihe 164399 China)
出处
《自动化技术与应用》
2023年第8期82-87,共6页
Techniques of Automation and Applications
关键词
深度长短记忆网络
网络数据流
异常检测
特征提取
deep and short memory network
network data flow
abnormal detection
feature extraction
