Abstract: Pressure-Preserved Coring Tools are critical to deep-sea resource assessment and development, and the closure performance of their sealing ball valve directly determines the success of acquiring in situ core samples. Addressing the difficulty of accurately quantifying the ball-valve actuating force under multi-variable interaction effects, this study integrates experiments with machine learning to achieve quantitative prediction across multiple factors. Through comprehensive factorial experiments, we obtain driving-force data under multi-level operating conditions involving temperature, seal-ring type, and lubricant viscosity, and systematically analyze main effects and interactions. The results indicate that lubricant viscosity is the primary influencing factor, exhibiting strong coupling with temperature. Building on these findings, we employ data-augmentation techniques and a gradient boosting regressor to construct a multi-factor prediction model. The model attains an R² exceeding 0.99, and quantitative feature-importance analysis yields a ranking consistent with experimental trends, validating the effectiveness of the approach. The results provide a reliable basis for the optimized design and materials selection of sealing systems in deep-sea PressurePreserved Coring Tools.