In the face of complex maritime traffic and dense logistics traffic flow, timely and effective tracking and prediction of ship trajectories is particularly important. The work aims to propose a method to solve the low accuracy and low efficiency of traditional ship trajectory prediction methods. A ship trajectory prediction model which combined the improved particle swarm optimization (IPSO) algorithm with the BP neural network was established based on AIS data. Historical ship trajectory data were used to predict future navigation trajectories. The historical ship trajectory data of Zhoushan Port in Ningbo was selected for the experiment, and the experimental results of the IPSO-BP model were compared with other models. Through comparing the results of different model trajectory predictions, it could be seen that the IPSO-BP model had good performance and high prediction accuracy, and is suitable for ship trajectory prediction. The use of IPSO-BP model can achieve more accurate ship trajectory prediction, which has an important guiding role for future ship hazard warning, ship anomaly monitoring, and other aspects.