A combined VMD-SSA-LSTM model is proposed to enhance the prediction accuracy of the ship prediction model. This will improve the compensation effect of the wave compensation system and resolve the hysteresis problem of the compensation system. The memory (LSTM) neural network (VMD-SSA-LSTM) decomposes the ship’s attitude signal using VMD and the sparrow search algorithm (SSA) to optimise the LSTM parameters. The combined VMD-SSA-LSTM model uses VMD and SSA to optimise the parameters of the LSTM network. This involves decomposing the ship’s attitude signal using VMD to obtain a finite set of intrinsic mode functions (IMFs). The IMFs are then obtained by decomposing the ship’s attitude signal using VMD. The LSTM is then optimised with SSA to predict each intrinsic mode function, and the prediction results are summed up to obtain the final prediction results. Real ship motion attitude signals measured at sea show that the combined model is more accurate than the LSTM, VMD-LSTM model.