Abstract: Due to the large volume and rich spatiotemporal features of 4D trajectory data, most prediction models have problems such as missing prediction dimension, insufficient extraction of spatiotemporal features, and short prediction length. To solve the above problems, a multi-layer convolutional bidirectional gated composite (MCBAC) model is established based on convolutional neural network (CNN) and bidirectional gated recurrent unit (BIGRU), introducing a multi-layer neural network strategy and a pre-processing algorithm for the adaptation of the trajectory data to realize the simultaneous processing of spatiotemporal features of the trajectory data and improve the 4D trajectory prediction accuracy. The experimental results show that the MCBAC model has significant advantages in terms of prediction accuracy, deviation range, and predictable duration when dealing with civil aviation 4D trajectory prediction problem, its error value is smaller than that of the comparison prediction model in any dimension, and the fit of the prediction results and the model performance are better.