Abstract: In this paper, we describe the characteristic of general regression neural network. Then the model of predicting the occurrence of liquefaction is proposed after considering the main factors. The supporting data is derived from the results of field tests of the two major earthquakes that took place in Turkey and Taiwan in 1999. After utilizing the multiple regression pretreatment to optimize the input data, the study uses the GRNN to implement the training and validation phase. The results have a higher degree of consistent with the actual situation and clearly demonstrate the capability of the proposed model to assess the liquefaction of soils. Then multiple linear regression is proved to improve the accuracy of forecasts as a good pretreatment method.