The kernel-LMS based network traffic prediction

Network traffic prediction(NTP) has become a very important technology in network traffic management and information security. The neural network based on MSE criteria is an key method of NTP. However, the nonlinear character of the network flow has led to certain limitation for the application of the method. Considering the nonlinear characteristics of the network traffic, the Kernel LMS(KLMS) algorithm is studied, and then the network traffic prediction mechanism based on KLMS algorithm is proposed, This mechanism can map the nonlinear data from low dimensional input space to high dimensional feature space through the kernel function to conduct a linear operation, making the calculation simple and effective. The simulation results show that compared with the LMS algorithm, KLMS has certain superiority in prediction precision.