Abstarct: Retaining walls are among the most important structures in civil engineering. Therefore, their safe design can lead to satisfactory results and economic efficiency in their design. Soil parameters are variable in soil mass. This means that the properties of the soil are not the same along the entire length of the soil laxyer and vary from point to point, which is called uncertainty. In reliability design, all these uncertainties can be considered and a safer design can be presented; While in conventional design methods, uncertainties are not taken into account and in fact the whole soil laxyer with the same characteristics is assumed. There are many ways to design the reliability of a system. In addition to all the advantages that these methods have in evaluating the reliability of a system, the existence of some problems such as high time cost and insufficient accuracy and problems such as these, led to the emergence of new methods. In this research, at the first, methods such as Monte Carlo sampling method and Latin hypercube have been evaluated and then to solve some problems of these methods, the combined methods of Monte Carlo and Latin hypercube with support vector machine have been used. In this method, first, using Monte Carlo sampling method and Latin hypercube, random samples are generated according to statistical parameters and input distribution function, and then using the support vector machine, the probability of structural failure is predicted. According to the obtained results, it was found that these methods, in addition to reducing the time cost of calculations, also have acceptable accuracy in structural design. Also, the effect of correlation and non-correlation between input parameters has been investigated. Finally, it was found that by considering the correlation coefficient, the reliability of the structure decreases and for a safer wall design, it is better to design by considering this parameter. The reliability coefficients in the regulations have been obtained through engineering experience and judgments, which in some cases can lead to changes or even breakdowns in the structure. In this dissertation, in the analysis of retaining walls, the interval of reliability of each of them has been obtained and the minimum and maximum values of reliability have been specified in two cases. In the first case, in terms of science, the reliability of the wall has not reached the desired safety range (reliability coefficient of 2.5-3) and in the second case, by making changes in the geometry of the retaining wall and re-analyzing that wall, it reaches the safety range. By comparing the outputs, it can be seen that for an almost identical confidence interval, two states emerge: the state in which the wall is ruptured and the state in which the wall is not ruptured. This indicates a more accurate and safer design of the reliability method than traditional methods. Also, in the end, a suitable reliability coefficient is proposed for each wall.