Considering the main factors with important influence on sand seismic liquefaction, the support vector machine （SVM） model is established, which includes seven indexes such as earthquake magnitude, SPT counts, relative density, soil layer depth, time history of earthquake, peak ground acceleration and epicenter distance. Taking surving data as samples for training and learn- ing, some functions are obtained in identification of sand sample. It is shown that the identification model of SVM analysis is an effective method to predict sand liquefaction with high prediction accuracy and could be used in practice.