基于FARIMA模型的网络流量预测

Network Traffic Forecast Based on FARIMA Model

论文基于长相关特性的时间序列分析方法,采用FARIMA模型对网络自相似业务进行研究,利用“后向预报”技术对序列进行分形反滤波,在模型辩识、参数估计中利用粗、精估计结合的方法建立模型。选择伯克力实验室的经典实测数据,利用FARIMA模型进行H值估计、分数差分定阶及消除长相关性的操作,实验证明了模型的有效性。

In this paper,the authors study network self-similar service based on FARIMA model by time sequence analysis with long range correlation.Backward forecast technique is used to distribute reverse filter wave and model founded with combination of coarse and extract in model recognition and parameter estimation.Traditional measure data from Berkeley Lab is selected and FARIMA model is adopted in H value estimation,fraction difference fix-step and remove Long- range correlation.Experiment result shows validity of the model.