高斯过程回归补偿ARIMA的网络流量预测

为提高网络流量时间序列的中期预测精度,提出一种高斯过程回归模型补偿自回归积分滑动平均(ARIMA)模型的网络流量预测模型.首先通过Brock-Dechert-Scheinkman统计量检验方法确定网络流量时间序列包含线性特征与非线性特征;然后利用ARIMA模型对网络流量时间序列进行非平稳建模,得到符合网络流量序列线性变化规律的模型,并通过人工蜂群算法优化的高斯过程回归模型对具有非线性特性的预测误差序列进行建模与预测;最后将ARIMA模型的预测值与高斯过程回归模型的预测误差值进行相加得到最终的网络流量预测值.仿真对比实验结果表明,提出的预测方法具有更高的预测精度和更小的预测误差.

Network Traffic Prediction Based on ARIMA with Gaussian Process Regression Compensation

In order to improve the medium-term prediction accuracy of network traffic, a network traffic prediction method based on auto regressive integrated moving average ( ARIMA ) with Gaussian process regression compensation was proposed. Firstly, the linear and nonlinear characteristics of network traffic can be determined by Brock-Dechert-Scheinkman statistics. Then, ARIMA model was used for modeling the non-stationary network traffic time series. The linear model of network traffic sequence was obtained. The artificial bee colony algorithm optimized Gaussian process regression model was used as the prediction model of predictive error sequences with the nonlinear characteristic. Finally, the final prediction value was obtained by adding predictive values of ARIMA model and predictive error values of Gaussian process regression model. Simulation comparison shows that the proposed prediction method has higher prediction accuracy with the smaller prediction error.