<p>Assessing slope stability under heavy rainfall is crucial for mitigating the risks of rainfall-induced landslides. This study proposes a machine learning-based framework for the rapid assessment of two-dimensional (2D) slope stability under heavy rainfall. To address the limitations of traditional numerical simulations in regional-scale evaluations, a predictive relationship was established between slope geometry, groundwater conditions, and the factor of safety (FoS). Two models, a hybrid grey wolf optimizer-support vector regression (GWO-SVR) model and a multilayer perceptron (MLP), were developed using geometric and hydrological features derived from a three-dimensional (3D) slope model. The models were trained with key input variables, including maximum and average slope gradients, as well as a simplified 2D representation of the groundwater table extracted from a continuous 3D groundwater surface. A case study of a highway slope in Nagasaki, Japan, revealed that 2D FoS values can be either higher or lower than their corresponding 3D results. This challenges the common assumption that 2D analyses are conservative compared with 3D analyses. The results showed that the GWO-SVR model outperformed the MLP model in both accuracy and computational efficiency. Furthermore, SHAP analysis enhanced the interpretability of the GWO-SVR model. It offers insights into the relative importance of the input features. Overall, the proposed framework has the potential to serve as a practical and scalable approach for evaluating slope stability across regional scales under heavy rainfall.</p>