基于自适应自回归模型的非线性流量负荷预测

The Adaptive Auto Regressive Model for Nonlinear Traffic Load Forecasting

针对新一代移动数据业务(MMS,KJAV,WAP)具有复杂的非线性特性和不平稳特性,采用鲁棒Kalman滤波算法,提出了一种自适应自回归滑动平均模型(AARMA),并将其应用于移动数据业务负荷预测中。实际预测结果表明,即使是对变动大且不稳定的移动业务流量,自适应ARMA模型稳定,预测精度高,且预测误差的白噪声特性明显。

Through robust Kalman filtering method,an AARMA（Adaptive Auto Regressive Moving Average） model is proposed to forecast the nonlinear and non-stationary data traffic of MMS,KJAVA and WAP of China Mobile.The forecasting results show that the AARMA model need less model parameters and has higher forecasting accuracy,and the forecasting error has the obvious characteristic of white noise.