In order to explore the complex non-linear relationship between concrete dam deformation monitoring data and various influencing factors, and to improve the deformation prediction accuracy of ultra-high arch dams, based on the twin support vector machine (TWSVM) model and the adaptive particle swarm optimization (APSO) method for parameter optimization, an APSO-TWSVM prediction model for the deformation of a very high arch dam was constructed by introducing position and velocity factors. Case analysis shows that the model can effectively explore the complex non-linear relationship between the deformation of arch dams and the influencing factors, and compared with the traditional SVM method, the model speed and accuracy have been significantly improved.