LSSVM and hybrid particle swarm optimization for ship motion prediction

Ship motion prediction is essential for the safety of shipboard helicopter. If roll/pitch/heave exceeds some prescribed operating limit, potential crashes may occur. In order to prolong the prediction length, a hybrid algorithm based on particle swarm optimization and simulated annealing (HPSO) is proposed to choose the parameters of least square support vector machine (LSSVM). The HPSO-LSSVM method is based on the minimum structure risk of SVM and the globally optimizing ability of HPSO. It is applied to solve the problems of nonlinear chaos time series prediction and ship motion prediction. Experimental results show that the proposed algorithm can escape from the blindness of man-made choice of the LSSVM parameters and enhance the efficiency of online forecasting.