基于改进型小数据量法的局域网流量预测

LAN traffic flow prediction using improved small data method

基于Takens理论对混沌时间序列进行相空间重构,对小数据量法进行如下改进:利用C-C算法计算嵌入维和延迟时间;以功率对频率加权并采用求平均的方法计算平均周期,使小数据量法更加完善。使用改进前、后的小数据量法分别仿真计算Lorenz系统混沌时间序列的Lyapunov指数并预测混沌时间序列,并计算实测局域网流量时间序列的最大Lyapunov指数并预测局域网流量时间序列。仿真及实验结果均表明,采用改进型小数据量法进行流量预测,精度更高、速度更快、预测点数更多。

First,the LAN traffic flow time series are reconstructed in the phase space using Takens theory.Then the embedding dimension and delay time are calculated via the C-C algorithm.Third,the average period is calculated via the frequency weighting derived from the power and averaging method.With the above steps the improved small data method becomes more complete.The improved small data method is applied to calculate the largest Lyapunov exponent of the chaos time series of the Lorenz system and the prediction data of the real measured LAN traffic flow time series.Results show that the LAN peak traffic flow is chaotic and the prediction based on the improved small data method is more accurate,faster,and more points can be predicted.