Abstract: Flight trajectory data has the problems of uneven sampling time and inconsistent spatial dimension. Moreover, the existing trajectory prediction methods are mainly oriented to ground traffic trajectory such as pedestrians and vehicles, and there are few flight trajectory prediction methods applicable to three-dimensional space. To address the above problems, this paper proposes a prediction model for 3D flight trajectory based on generative adversarial network. The model resampled the flight trajectory data to unify the sampling time interval and eliminate the influence of extreme variation between different spatial dimension. The time series characteristics in the data and the interaction information between different targets were used to generate the predicted trajectory. Experiments show that compared with traditional trajectory prediction methods, the proposed model reduces the ADE by more than 29%, which verifies the effectiveness of the model in the prediction of flight trajectory in 3D space.