There are often high and steep slopes during the construction process of large projections, whose deformation often leads to geological disasters. This brings harm to the safety of people"s life and property. Hence, collecting displacement data efficiently and establishing a suitable and accurate hybrid prediction model becomes essential. This study aims to propose the Trans-CNN hybrid model by fusing the CNN convolutional layer and residual residual structure in the Transformer model. This optimized Transformer algorithm model was employed for the displacement data collected by the Beidou satellite system in a large water conservancy project in Chongqing. The results found that the mae, mse and rmse values of the Trans-CNN model are lower than single models, which proves the prediction accuracy of the Trans-CNN. Thus, it can provide a feasible scheme for the prediction and analysis of the deformation of the high and steep slope in the implementation of other projects.