Research on Slope Stability Prediction Based on MC-BKA-MLP Mixed Model

Quantifying slope mechanical parameters as comprehensive indicators is crucial for predicting slope stability. The Mohr–Coulomb (M-C) criterion, a classical method for determining the relevant parameters of rock mass mechanics, effectively reflects the failure characteristics of rock masses in most types of slopes. Based on this, effective stress and shear strength from the M-C criterion are selected as key indicators, and a characteristic dataset is constructed by integrating these with other influencing factors of slope stability. The safety factor, calculated using the Bishop method within the framework of limit equilibrium analysis, serves as the output variable. Subsequently, a novel Black Kite Algorithm (BKA) was developed to enhance the prediction model of a multilevel perceptron neural network. The results demonstrate that the mean square error (RMSE) of the BKA-MLP model is merely 2.41%, significantly lower than that of alternative models. Additionally, the R2 value reaches approximately 95%, indicating a high level of interpretability. The SHAP-based interpretability analysis of the trained model highlights effective stress, shear strength, and slope angle as the three most sensitive features. Based on these findings, targeted landslide prevention measures were proposed, providing a new approach for slope stability analysis and disaster prevention.