Structural material parameters will directly affect the stability and deformation of roller compacted concrete (RCC) dams. To analyze the influence degree of structural material parameters of RCC dams on the safety evaluation indexes, this study proposes a global sensitivity analysis method based on extreme learningmachines optimized by an improved sparrow search algorithm with Sobol method (ISSA-ELM-Sobol). First, the safety evaluation indexes of the RCC dam and their influencing factors are determined; second, a reasonable finite element model is established; third, the sample set is generated using the Latin hypercube sampling technique; fourth, the ISSA-ELM model is established to replace the finite element calculation; finally, based on the established ISSA-ELM model, the Sobol method is used for global sensitivity analysis. A case study for a typical RCC dam in Sichuan province of China showed that the material properties of the foundation and the cohesion of potential sliding surfaces have the most obvious influence on the safety of RCC dams; the ISSA proposed in this study can overcome the problem that sparrow search algorithm is easy to mature prematurely and fall into a local optimum. The ISSA-ELM model established in this study has significant advantages in solving nonlinear regression problems.