As the Earth-rockfill dam deformation has strong nonlinear characteristics, and the traditional statistical model is often not accurate and the error is big, the support vector machine (SVM) model is introduced. The key parameters penalty factor C and kernel function parameter σ of support vector machine are optimized by artificial bee colony algorithm (ABC) to improve the fitting and prediction accuracy, and the ABC-SVM model is established to monitor the deformation of earth-rockfill dam. The experiment shows that the ABC-SVM model has high prediction accuracy and strong generalization ability compared with the traditional multiple regression model and SVM model. It has a good effect to use ABC-SVM model to predict the deformation of earth-rock dam, and the model can be used in the field of dam safety monitoring.