Abstract: Reliable mechanical parameters of surrounding rock are imperative for the accurate prediction of tunnel deformation. A GA-BP based neural network back analysis system is proposed and automatic searching for BP network parameters can be realized, with the efficiency of inversion analysis increasing greatly. This GA-BP intelligent back analysis system is applied to the inversion of rock mass parameters and deformation prediction for the Dadingshan tunnel passing underneath the Shenyang-Dandong expressway. The results show that the convergence speed of the GA-BP back analysis system is fast and the parameter inversion of the surrounding rock and deformation prediction is accurate.