Abstract: To reduce the cost of soil-nailed wall of deep foundation pit and ensure its safety, a model was developed for its parameter optimization, including the number of lines of nails, their horizontal/vertical space, diameter and length. Traditional methods are apt to get locally optimization solution when used for complex optimum design problems like the parameters of soil-nailed wall. In this respect the genetic algorithm (GA) is a method available to get global optimum, but it often has such shortcomings as premature convergence and oscillation in iterative process. Therefor, a forward-and-backward searching algorithm is proposed in combination with GA and some measures for improvement, thus forming an improved hybrid genetic algorithm (IHGA) which, as proved by an engineering example, offers an optimum design that not only provides the stability for deep foundation pit but also reduces the cost of engineering materials.