Abstarct: The main basis in the analysis of stability and design of underground structures is the knowledge of the geomechanical parameters of the underground spaces. There are various methods for determining these parameters, which are sometimes used and only take a lot of time. One of them is back analysis methods. Various methods have been proposed for back analysis. The elementary methods include Direct, inverse and statistical methods. These methods take a lot of time and of course are not very accurate. The inverse method is simplification assumptions such as uniform environment, linear stress and single stage drilling. In reality, the situation is not like this. In the direct method, the optimization of the objective function requires multiple corrections of parameters, each of wich requires at least on calculation. Numerical solution is therefore time consuming, especially for large models. Recently, intelligent Back analysis methods have been proposed to overcome these problems. The case study in this project is Alborz tunnel. Alborz Tunnel is a tunnel in Tehran-Shomal subway. In this study, two types of back analysis systems have been selected to determine the geomechanical parameters of rock mass and in-situ stress conditions for three stations in different formations. Considering that the type of major formations in Alborz tunnel is of consolidation type C, this type has been used in this study. For this purpose, using the monitoring data and geological studies, numerical modeling of the Alborz tunnel by FLAC3D was performed to produce the input-output data pairs required for the design of the intelligent back analysis system. Then, two types of back analysis systems PSO and SVM were selected for this study after searching and optimization, they determined the parameters of the rock mass. Using sensitivity analysis, the values of modulus of elasticity, stress ratio and Hoek-Brown criteria model parameters were selected as input parameters. The use of the Hoek-Brown criteria model was chosen as the criteria model in this study, because it is closer to the actual results and more inputs are used to determined them. To evaluate the performance of the selected intelligent systems, after estimating the values of the rock massparameters, the data were compared with the instrumentation results of the sections. The results showed that the selected systems correctly predicted defferent values of rock mass. Using the error calculated for each of the systems, it was found that the PSO algorithm had an error of 1/7% and the SVM algorithm had an error of 3/75%. Finally, it was shown that the PSO algorithm has a very good performance in estimating the input parameters of the back analysis system.