To make the prediction result of dam safety more accurate with support vector machine(SVM), a genetic algorithm-support vector machine(GA-SVM)dam safety prediction model was proposed based on the genetic algorithm optimization. With the minimum error of k-CV verification as the optimization target, the genetic algorithm was introduced to optimize the penalty parameter c and kernel function parameter g of SVM. The model took the influence factor as the input and the effect quantity as the output. Then the training sample data were used to train the support vector machine, and the trained model was used to predict the effect quantity. According to the 3σ criterion in the probability and statistics theory, the three-grade index and discriminant criterion of dam safety were established. Taking an actual large-reservoir dam as the case, the GA-SVM model of this dam was established and compared with the SVM model and stepwise regression model. The prediction results show that the predicted value of GA-SVM model is closest to the measured value, and its prediction accuracy is about three times higher than both SVM model and stepwise regression model. Therefore, the GA-SVM prediction model has a good prediction precision and can be used for the dam safety prediction.