Research Of Network Traffic Prediction Method Based On Hierarchical Echo State Network With Adaptive Reservoir

Network traffic is an important parameter to evaluate network security.Network traffic prediction can foresee abnormal network traffic,which is beneficial to network attack prevention and network congestion control.Therefore,network traffic prediction is a hot research topic in the field of network security.Echo state network(ESN)is a new neural network with strong nonlinear processing capacity and short-term memory capacity,and it has fast training speed.ESN can handle the sequentiality and nonlinearity of data,but it still has some problems when it is applied to network traffic prediction.First,the diversity of network traffic means that different network traffic has various characteristics.As the core of ESN,the reservoir will be fixed rather than adjustable once it is generated,which limits the prediction performance of ESN in different network traffic.Second,the multi-scales of network traffic means that the same network traffic displays different laws of data change on multiple time scales.If the multi-scales can be utilized,the performance of network traffic prediction will be further improved.However,ESN cannot effectively utilize multi-scales.In view of the above problems,this thesis studies a new network traffic prediction method based on hierarchical ESN with adaptive reservoir.The details are as follows:1.To achieve universal excellent prediction performance in different network traffic,this thesis proposes a network traffic prediction method based on ESN with adaptive reservoir(ESN-AR).First,the network traffic prediction model based on ESN-AR is constructed,in which the idea of generative adversarial network is incorporated into ESN to adaptively adjust the reservoir.Specifically,ESN is used as the generative model to predict network traffic,and feedforward neural network is used as the discriminative model to distinguish between the real network traffic and the predicted network traffic.Second,the adversarial training algorithm of ESN-AR is proposed to obtain the appropriate reservoir depending on the network traffic characteristics.Finally,ESN-AR is applied to the prediction of three actual network traffic with different characteristics.Simulation results show that compared with the state-of-the-art models,the proposed method achieves more accurate and stable prediction performance.2.To further improve the prediction performance in network traffic with multi-scales,this thesis proposes a network traffic prediction method based on hierarchical ESN(HESN).First,the network traffic prediction model based on HESN is constructed,in which the hierarchical architecture is combined with ESN-AR to utilize the multi-scales of network traffic.Specifically,the multi-scale network traffic representation method is used to represent network traffic hierarchically,and ESN-AR is used in each layer to predict network traffic on the corresponding time scale.Second,the training algorithm of HESN is proposed to train ESN-AR of each layer from the bottom up.Finally,HESN is applied to the prediction of three actual network traffic.Simulation results show that compared with ESN-AR,the proposed method further improves the prediction performance.