Abstract: ‍ ‍In the terminal area， the airspace environment is complex and the flights are dense. Accurate trajectory prediction can greatly improve air traffic service level， and ensure aviation safety. To solve the problem of multi flight and high-precision 4D trajectory prediction required by the terminal area， an algorithm that combines Density-Based Spatial Clustering of Applications with Noise （DBSCAN） and Gated Recurrent Unit （GRU） is proposed. Through DBSCAN， the flights with similar trajectory in the terminal area are clustered into a cluster， and then the GRU is used to train the trajectory prediction model for the trajectories of different clusters. When a flight enters the terminal area and needs to be predicted， first， the flight is judged which cluster belongs to， and then the trajectory prediction model corresponding to this cluster is used for 4D trajectory prediction. Compared with the traditional prediction method that only studies a single flight， this algorithm effectively uses the trajectory data in the terminal area. The built model can predict the trajectory of multiple flights， expand the scope of application of the model， and improve the prediction accuracy of the trajectory prediction.