Hybrid random forest models optimized by Sparrow search algorithm (SSA) and Harris hawk optimization algorithm (HHO) for slope stability prediction

Abstract

The intelligent prediction of slope stability is crucial for slope management and aboveground structure design. Due to the complexity and nonlinearity of slope stability evaluation, intelligent prediction models often perform better. This study developed two optimizers, the Sparrow search algorithm (SSA) and the Harris hawk optimization algorithm (HHO), to optimize the random forest (RF) model for evaluating slope stability. A database of 444 slope cases was established. The data was split into 80% for training and 20% for testing. The optimizers were used to determine the hyperparameters of the models. Four classical classification models were introduced for comparison. Five-fold cross-validation was employed, and model performance was evaluated using accuracy, precision, recall, and F1-score. A comprehensive method was developed for evaluating the performance of models. The SSA-RF model exhibited the best performance with the highest accuracy (0.9101), precision (0.9500), recall (0.8636), F1-score (0.9048), and AUC (0.9407). The Gini coefficient indicated that unit weight was the most influential indicator (41.16). These methods offer a promising approach for predicting slope stability, applicable in slope engineering practice.