Abstract: This paper aims to present a back-analysis method on parameter of soil nonlinear elastic constitutive models. . The study is based on the coincident optimization theory of the genetic algorithm, and the commonly used foundation loading experiment. A parameter back analysis method is established on the nonlinear elastic constitutive models of soils by means of joining together the forward computation and the genetic optimization algorithm. Meanwhile, the data from the foundation load experiment of loess stratum are used. To implement the progress of the parameter back-analysis for the nonlinear elastic constitutive model of loess stratum. The results show that the proposed method can accomplish the coincident optimization of much relative parameters in the nonlinear elastic constitutive model of soil. It does give a good estimation without imposing any stringent requirements on the initial values of parameters. Therefore, for analyzing the deformation characteristics and the interaction characteristic of soils and the structures, the paper provides a good method to define the constitutive model parameters of the soils.