Voltage sags are one of the primary factors in power quality issues that lead to losses for sensitive users and reduce the operation resilience of distribution networks. However, due to the lack of accessibility in sensitive users' production information, accurately quantifying the resilience of distribution networks under the impact of voltage sags is challenging. In this letter, we first define an operation resilience index using a trapezoidal curve. Considering the varying tolerance levels of sensitive users to voltage sags, a feature indices system is estabilished using the adaptive S-transform, and a sample dataset is generated through the Monte Carlo method. Finally, we establish a mapping relationship between sag characteristics and operation resilience indices using the XGBoost-Stacking algorithm. This data-physics hybrid-driven model offers a quantitative approach for developing resilience enhancement strategies.