关键词: 高斯过程, 遗传算法, 延迟时间, 网络流量, 嵌入维数

Abstract: In order to improve the prediction precision of network traffic, this paper proposes a network traffic prediction model based on combined kernel function Gauss Process（GP） to describe the nonlinear and time-varying characteristics of network traffic. Firstly, the time delay and embedding dimension of network traffic are calculated by self correlation method and false nearest neighbor method, and training samples of network traffic are generated, and then the training set is input to combination kernel function GP learning to establish a network traffic prediction model which the genetic algorithm is used to find the optimal parameters of GP, and finally, the simulation experiments is carried out on network traffic data. The results show that, compared with the other models, the proposed model can obtain higher prediction precision of network traffic, the prediction results are more stable and reliable, so it has great practical application value.

Key words: Gaussian Process（GP）, genetic algorithm, delay time, network traffic, embedding dimension