自回归与神经网络组合的网络流量预测模型

Network Traffic Prediction Based on Hybrid Model of Autoregressive and Neural Networking

网络流量具有时变性和非线性,单一预测方法难以准确描述网络流量变化规律,为提高网络流量预测准确率,提出一种网络流量组合预测模型(ARIMA-BPNN);首先采用ARIMA对网络流量进行预测,然后采用BPNN对网络流量非线性变化规律进行预测,且遗传算法优化BPNN初始权值,最后将两者预测结果作为BPNN输入进行二次预测,得到ARIMA-BPNN预测结果;仿真实验结果表明,相对于ARIMA、BPNN,ARIMA-BPNN提高网络流量预测精度,在网络管理中有着广泛的应用前景。

Network traffic with time--varying and nonlinear, a single prediction method is difficult to accurately describe the network flow variation, in order to improve the prediction accuracy of network traffic, put forward a kind of hybrid prediction model （ARIMA-- BPNN ）. The ARIMA of network traffic prediction, and then the BPNN on network traffic prediction of nonlinear changes, and genetic algorithm optimization BPNN initial weights, finally, the forecasting results between BPNN input as two prediction, forecasting results by ARI MA--BPNN. The simulation results show that, compared with ARIMA, BPNN, ARIMA--BPNN increase network traffic prediction accuracy, in network management and has extensive application prospect.