Initial geo-stress is one of the most important factors in geotechnical designing and constructing. Through effective numerical analysis method, and on the basis of in situ measurements, a precise geo-stress field can be gotten which can be used in whole large scale area. On the basis of the geological mechanics analysis, the geo-stress is mainly caused by gravity and the geological constructional stress field. The numerical model used to analyze the geo-stress field is established on the basic of such a view, and multivariate regression analysis method is used for fitting analysis. As engineering application, Xue Feng Shan tunnel is analyzed which is a very long tunnel located at Shangrui Highway, Hunan Province. It is about 2*7500 m long and will be the longest highway tunnel that already in construction designing step in China. The maximum depth of the tunnel is about 850 m. The geo-stress' magnitude and distribution will play a very important role on the stability of tunnel when carrying out tunnel excavation under such a situation. In situ geo-stress measurements using hydraulic fracturing method have been carried out in three drill bores, altogether 28 measure points. To simulate the geo-stress field, the computational area is set 8000 m*2000 m wide. Based on the geological information of XUE FENG SHAN tunnel, the main rock types considered include met sandstone, silica sand shale and mylonite, and the mainly considered faults include F2, F3, F4, F5 and F6. Using the information of 28 measure points in three holes, as well as four three-dimensional FEM simulations of geo-stress, through carrying out multiple regression analysis, we get the regression coefficient of four independent variable (corresponding gravity stress, constructional stress on the axis direction in horizontal plane, and shear stress in horizontal plane) as L1 = 0.9737784, L2 = 28.33104, L3 = 4.792041, L4 = 15388.15. The multiple correlation coefficient r = 0.9542671, closing to 1, which means the regression result is accurate. After the four independent variables are gotten, then taking them into the first equation, the regression value at the measured points can be obtained. (A). Reprinted with permission from Elsevier. For the covering abstract see ITRD E124500.