Abstract: Roof water hazard poses a serious threat to mine production safety, and its occurrence is closely related to the deformation and failure of overlying strata. Physical similar simulation experiments serve as a crucial method for studying this issue, where material ratios adhering to similarity laws are key to ensuring the scientific validity of experimental results. Uniaxial compressive strength is one of the most basic and key mechanical indexes in the design of similar materials. Traditional research predominantly relies on orthogonal experiments, which struggle to reveal nonlinear interactions among independent variables and lack the capability for continuous prediction of material properties. Standard specimens of ø50 mm × 100 mm were prepared using river sand and barite powder as aggregates, gypsum and cement as composite cementing materials, and water as the solvent. Employing the Response Surface Methodology (RSM), with aggregate-binder ratio (4∶1−6∶1), cement-binder ratio (3∶10−7∶10), and barite powder content (10%−30%) as independent variables, a quadratic regression model was constructed based on 17 sets of uniaxial compressive strength tests and Box–Behnken design, and its accuracy was validated through five uniaxial compression tests. The results indicate that the established model has a coefficient of determination R² = 0.9908, with a difference of less than 0.2 between the adjusted R² and predicted R², and a signal-to-noise ratio of 32.74. The influence of barite powder content on uniaxial compressive strength exhibits a “U-shaped” trend, showing strong interactions with the aggregate-binder ratio and cement-binder ratio. The validation results showed a relative error of ≤7%. The findings provide a basis for the mix design of similar simulation materials for overlying strata failure, significantly enhancing the reliability of physical simulation experiments and offering theoretical guidance for studying the failure mechanisms of overlying strata.