基于HP滤波的ARMA-ABCSVR-GABP网络流量预测

ARMA-ABCSVR-GABP NETWORK TRAFFIC PREDICTION BASED ON HP FILTER

针对当前网络流量无法根据流量变化的特征进行预测,且通过单一或者组合模型依然得不到较高准确率的问题,提出一种基于HP(High-Pass Fliter)滤波的流量预测模型。基于高铁站流量数据日高夜低的周期特性以及流量波动增长的长期趋势,依据HP滤波将网络流量分解成周期序列及趋势序列。利用自回归-滑动平均模型(ARMA)对平稳序列预测的优势来进行周期变化的预测;使用人工蜂群算法(ABC)优化后的支持向量回归机(SVR)对趋势序列进行预测;将二者预测的结果叠加,使用遗传算法优化的BP神经网络(GABP)进行结合预测,进一步提高准确率。结果显示,该预测方法可靠,较其他方法具有优越性。

In order to solve the problem that the current network traffic cannot be predicted according to the characteristics of traffic changes,and the high accuracy still cannot be obtained through a single or combined model,a traffic prediction model based on HP filtering is proposed.Based on the periodic characteristics of high daytime and low nighttime traffic data in high-speed railway station and the long-term trend of traffic fluctuation growth,the network traffic was decomposed into periodic sequence and trend sequence according to HP filtering.The advantage of ARMA in predicting stationary sequence was used to predict periodic change.The SVR optimized by ABC was used to predict trend sequence.Finally,the prediction results of the two were superimposed,and GABP was used to further improve the accuracy.The results show that the prediction method is reliable.Compared with other methods,the superiority of this method is proved.