Abstract: In order to ensure the foundation pit construction safety, displacement monitoring was carried on the foundation pit soil. At the same time, according to the measured displacement, the displacement parameters inversion algorithm was taken to calculate the foundation pit soil mechanical parameters, then calculate the stability of foundation pit soil. According to the basic calculation method of genetic algorithm optimizing neural network, in this paper, use uniform design method to design the samples of the foundation pit soil displacement and mechanical parameters. The finite element analysis was used to calculate soil displacement with different parameters. The GA-BP neural network was trained to describe the sophisticated nonlinear relationship between displacement and mechanical parameters of the foundation pit soil. Finally, the actual displacement was input into the trained GA-BP neural network to obtain the soil mechanical parameters. As an example, the soil cohesion, friction angle, elastic modulus and Poisson ratio of Changchun West Railway Station foundation pit were back-analysis by the above method. The calculation displacement obtained by finite element analysis with inversion results and compared with measured data. The comparison result shows that GA-BP neural network has high precision in excavation soil parameters inversion which can meet the needs of engineering.