Combination kernel function least squares support vector machine for chaotic time series prediction

Abstract

Considering the problem that least squares support vector machine prediction model with single kernel function cannot significantly improve the prediction accuracy of chaotic time series, a combination kernel function least squares support vector machine prediction model is proposed. The model uses a polynomial function and radial basis function to construct the kernel function of least squares support vector machine. An improved genetic algorithm with better convergence speed and precision is proposed for parameter optimization of prediction model. The simulation experimental results of Lorenz, Mackey-Glass, Sunspot-Runoff in the Yellow River and chaotic network traffic time series demonstrate the effectiveness and characteristics of the proposed model.