Considering some defects of BP Neural Network, this paper proposes a GA-BP （ Genetic Neural Network） learning algorithm, which is obtained by using Genetic Algorithm to optimize neural network trained by BP algorithm, and the GA-BP model of sand liquefaction is established. Based on the observation data of sand liquefaction, the computing results of the GA-BP model and the BP model prove that the former is nicer than the later in predicting the sand liquefaction.