In-situ stress is an important factor in stability analysis of slope. Based on the measured stress data in the high and steep slope of Xingshan iron open pit

the inversion of in-situ stress field in the slope is performed by using multivariate linear regression method considering the impact of the gravity of rock mass and tectonic stress

in which 11 kinds of boundary conditions are included in five categories. In terms of the limits of linear and the whole region inversion

this method cannot accurately inverse the stress field of the metal ore body

which has extremely complex internal structure and big differences in lithology and topography within the mining area. In order to overcome the above defects

nonlinear neural network inversion method is used. According to the differences in lithology and topography of the rock mass

the mine is divided into five zones. Sub-regional inversion of the in-situ stress field of the mine is achieved by introducing a lateral pressure coefficient k0. The study results show that under a complex geological condition

the inversion of initial in-situ stress by neural network method can give a rise to a more reasonable result.