Abstract: Due to the resource limitation and topology change in satellite network, the article puts forward higher requirements for the accuracy and efficiency of the network traffic prediction algorithm, and the traditional prediction model is no longer suitable for the satellite network. The author presents a kernel least mean square algorithm (KLMS) with adaptive step length and adaptive kernel width, namely AKLMS, which maps the nonlinear data from low dimensional input space to high dimensional feature space through kernel function, and the algorithm will adaptively adjust the step length and kernel width based on the instantaneous error in the iterative process. Simulations show that the AKLMS algorithm has great improvement on the convergence speed and prediction accuracy of the flow compared with the KLMS and least mean square (LMS), which will provide strong decision support for traffic planning and routing design in satellite network.

Key words: satellite network, kernel least mean square, adaptive step length, adaptive kernel width, network traffic prediction