In the present study, a CFD-based database is proposed to predict surficial pressure loading. To evaluate flow properties, a three-dimensional multi-phase Navier-Stokes flow solver based on unstructured meshes was utilized. Relative motion between a platform and a projectile was described using six degrees of freedom (6-DOF) equations of motion and an overset mesh technique. Accuracy of the 6-DOF module in the flow solver was validated by solving a benchmark problem. Four major parameters related to the surficial pressure loading were selected, which are escaping velocity, pressure of compressed gas flows, freestream velocity, and depth of water. Initial sample points were estimated by applying an improved Latin hypercube sampling method. To evaluate a flow condition with the maximum surficial pressure loading in a given database, a genetic algorithm and an artificial neural network were utilized. Then, accuracy of the database was iteratively evaluated by comparing the maximum pressure loading with those by the CFD prediction. It was found that the relative errors are less than 0.1% for both the projectile and the platform with 24 resultant sample points. It was also found that the escaping velocity and the freestream velocity significantly affect the characteristics of the surficial pressure loading due to the rotational disturbance around the projectile.