A Hybrid Particle Swarm Optimization Algorithm for Predicting the Chaotic Time Series

A novel hybrid particle swarm optimization (HPSO) is proposed, which the gradient descent learning algorithm is combined with modified particle swarm optimization (MPSO). Firstly, the MPSO was determined by linearly decreasing inertia weight and constriction factor weight to speed up global search, also crossover and mutation operation was embedded to avoid the common defect of premature convergence. Furthermore, gradient descent learning algorithm searched for the model parameter of radical basis function neural networks (RBFNN) to speed up the local search. Using the proposed HPSO algorithm based on RBFNN, we simulated the chaotic time series prediction of Henon map to test the validity. Simulation results show that the HPSO can accurately predict chaotic time series. lt provides an attractive approach to study the properties of complex nonlinear dynamic system