Abstract: In order to improve the network traffic forecasting accuracy, this paper proposes a network traffic forecasting model based on phase space reconstruction and neural network optimized by CPSO algorithm（CPSO-BPNN）. The parameters of BP neural network, delay time and the embedding dimension are optimized by Chaos Particle Swarm Optimization algorithm, and the data of network traffic are reconstructed. BP neural network is used to train to establish network traffic forecasting model based on the optimal parameters, and the simulation experiments are carried out to test the performance of network traffic forecasting model. The simulation results show that the proposed model can describe the change trend of network traffic, and improve the network traffic forecasting accuracy.

Key words: network traffic, forecasting accuracy, phase space reconstruction, neural network

摘要： 为了提高网络流量的预测精度，提出了一种混沌粒子群算法优化相空间重构和神经网络的网络流量预测模型（CPSO-BPNN）。利用混沌粒子群算法对BP神经网络初始参数、延迟时间、嵌入维数进行优化，根据延迟时间、嵌入维数对网络流量数据进行重构，BP神经网络根据初始参数进行训练建立网络流量预测模型，通过仿真实验对模型性能进行测试。结果表明，CPSO-BPNN可以准确描述网络流量的复杂变化趋势，提高了网络流量的预测精度。

关键词: 网络流量, 预测精度, 相空间重构, 神经网络