The accelerated degradation of pollutants in the vacuum ultraviolet/persulfate (VUV/PS) system is often attributed to the effective excitation of H2O and persulfate (PS) by UV185. However, the direct excitation of pollutants by UV185 has been largely overlooked, which may result in an underestimation of UV185's role in pollutant degradation. To address this gap, this study integrates machine learning and density functional theory (DFT) calculations to elucidate the mechanism of UV185 in the VUV/PS system through a data-driven approach. Initially, the ground-state and excited-state molecular descriptors for 30 types of organic pollutants were derived via DFT calculations and used as input parameters. Subsequently, a stochastic forest model was employed to predict the degradation kinetic constants and mineralization rates of pollutants in various systems, serving as output parameters. By evaluating the model's performance under different input conditions, molecular descriptors with high relevance to the output parameters were identified and retained. Ultimately, the most influential input parameters in each model were analyzed using the Shapley Additive Explanation (SHAP) method, which facilitated the speculation of the reaction mechanism. The findings revealed that, compared to the UV system, the contribution of S1 excited state descriptors and nucleophilic reaction-related descriptors in the VUV system was markedly enhanced, suggesting that UV185 promotes the transition of pollutants to a more reactive S1 state, thereby accelerating their degradation and mineralization through enhanced photolysis and nucleophilic reaction pathways.