For the prediction of seismic liquefaction of sand, Fisher discriminant model is established with eight factors listed as follows: mean diameter, coefficient of non-uniformity, blow number of standard penetration test, underground water depth, sand depth, ratio of shearing stress to effective overburden stress, seismic intensity, epicenteral distance. Through computing practical examples and assessing the model, the Fisher discriminant analysis model is manifested to be of rationality and reliability and the model identification results are consistent with the neural network output, and are better than the discriminant results with the specification method and Seed method.