基于混沌时间序列和神经网络的网络流量预测方法

网络流量时间序列具有复杂的非线性和不确定性特征，故提出以相空间重构理论与递归神经网络相结合的网络流量预测方法。以相空间重构理论确定最佳延迟时间和最小嵌入雏数，重构网络流量时间序列。将重构后的时间序列运用递归神经网络来训练，得到合适的模型，并用于网络节点中网络流量的预测。将该方法应用于实际数据预测，其结果与传统的时间序列预测方法结果相比较，提高了预测精度和稳定性，证明了该预测模型和方法在实际时间序列预测领域的有效性和实用性。

Approach of Network Flow Prediction Based on Chaotic Time Series and Neural Network

For the network flow time series has complex non-linear and uncertainty characters,a approach of network flow prediction was presented according to the phase space reconstruction theory(PSRT) combined with recurrent neural network(RNN).The optimal delay tmie and minimal embedding dimension are determined by PSRT,and then the network traffic time series is reconstructed.The reconstructed time series is trained by RNN to obtain a suitable model,which is applied to the prediction of network flow in the network...