This study presents a predictive model for assessing the stability of unsaturated natural slopes in mountainous regions affected by heavy rainfall. The proposed model is based on prediction functions capable of providing safety factors of unsaturated slopes quickly considering the geometric properties of the slope, geotechnical properties of the soil, and dynamic characteristics, such as soil moisture changes due to rainfall. Initially, a database is created that considers the safety factors of several soil slopes evaluated under different conditions through a Limit Equilibrium Method. Then, a multi-objective optimization method is performed using differential evolution algorithms to minimize the root mean square error (RMSE) and model complexity, resulting in a Pareto front that balances error and term count. Finally, the model’s performance is assessed using regression metrics such as RMSE, coefficient of determination ( R 2 ), and mean absolute percentage error (MAPE). A categorical classification approach is also adopted, where slopes are deemed stable if the safety factor exceeds 1.5. The model’s accuracy across different soil types averages 95.39%, with a recall rate of 95.99%, providing robust predictions of slope stability and offering valuable insights for geotechnical engineering and landslide risk management in regions prone to heavy rainfall-induced landslides.
Resende et al. (Mon,) studied this question.