Distributed optical fiber strain sensing technology can realize the full distributed monitoring of arch dam deformation, but the strain is directly measured instead of displacement. Therefore, it is necessary to study the displacement transfer of such complex structures as arch dams according to strain measurement. In this paper, Bayesian Optimized Long-short-time Memory Neural Network (BO-LSTM) and distributed optical fiber monitoring are innovatively combined to establish a three-dimensional model for measuring dam strain by optical fiber, which realizes the transformation between strain and displacement. The feasibility of different spatial resolutions is discussed by using BO-LSTM, and the strain input and displacement output are selected by using pure arch method. The results show that the strain–displacement transformation of arch dam can be realized accurately based on the BO-LSTM transformation model. The accuracy of the model is improved by screening strain values by analogy pure arch method.