Application of ANN Artificial Network in Slope Behavior Evaluation Using Machine Learning Technique

In geotechnical engineering, evaluating slope stability remains a significant challenge due to inherent soil variability and environmental uncertainty. This study explores the application of Artificial Neural Networks (ANN) in predicting the factor of safety (FoS) of slopes using dimensionless input variables. An initial dataset of 349 samples was refined to 259 validated cases after removing outliers and incomplete entries. The input parameters were transformed into nondimensional forms c / γh, (tan(⌀)) / (tan (β)), and p / γh based on fluid mechanics principles to improve generalization and reduce dimensional bias. A feedforward ANN model with a 3–7–1 architecture, trained using backpropagation and hyperbolic tangent activation, was developed and validated using 5-fold cross-validation. The ANN model achieved strong predictive performance with MAE = 0.27, RMSE = 0.46, and R2 = 0.946, outperforming comparative models including GMDH and CN2. The results indicate that ANN is a robust and reliable tool for slope safety prediction across variable geotechnical conditions.