CAJ | 학술논문

为克服传统时间序列分析方法对小数据信息数据和非平稳序列检测不稳定的限制，引入滑动窗口模型思想，提出了滑动时间窗口模型的网络流量序列重组空间异构的检测方法。通过计算仿真得到不同时间窗阈值下的网络流量序列递归图，检验出网络总出口流量的确定性。通过提取递归图中异常特征点的定量递归特征的方法实现对流量异常的检测和评估。仿真实验表明，提取的流量序列定量递归特征具有较强的稳定性和自相似性，算法能有效检测出网络流量序列的隐藏异常波，尤其适合于小数据量时间序列和非平稳数据的检测和分析。
위극복전통시간서렬분석방법대소수거신식수거화비평은서렬검측불은정적한제，인입활동창구모형사상，제출료활동시간창구모형적망락류량서렬중조공간이구적검측방법。통과계산방진득도불동시간창역치하적망락류량서렬체귀도，검험출망락총출구류량적학정성。통과제취체귀도중이상특정점적정량체귀특정적방법실현대류량이상적검측화평고。방진실험표명，제취적류량서렬정량체귀특정구유교강적은정성화자상사성，산법능유효검측출망락류량서렬적은장이상파，우기괄합우소수거량시간서렬화비평은수거적검측화분석。
In order to overcome unstable property of the traditional time series analysis method in the detection of small da-ta information data and non stationary sequence, an new and improved anomaly detection of network traffic sequence with space reconstruction based on sliding time windows model was proposed. The recurrence quantification analysis features such as recurrence rate and determinism, etc. were proposed for analyzing the stability and prediction property. And the anomaly detection and estimation of campus network traffic was realized based on recurrence plot analysis by extracting the abnormal points in recurrence plot. Simulation results show that the recurrence quantification analysis features of network traffic sequence have strong stability and self similarity, detection algorithm can detect the hidden abnormal wave effective-ly, it is suitable for small data time and non-stationary data series detection and analysis especially.