Considering that carrier-based aircrafts are affected by deck motion disturbances during landing, a prediction method based on the improved BP neural network (NN) using Grey Wolf (GW) algorithm is proposed. The model of deck motion is developed in this paper. Considering the irregular characteristics of deck motion, the BP NN is used to train it and predict the future deck motion information in advance. Aiming at the defects of BP such as slow convergence speed and converging local optimum, the GW algorithm is designed to optimize its initial threshold and weight to achieve quickly global convergence. The predicted information is used as the correction term of landing guidance command and the back-stepping method is used for control. Simulation results show that the proposed method can estimate and predict deck motion more effectively and enhance the control accuracy of landing trajectory tracking.