Research on pile bearing capacity prediction improved by grey wolf optimization in the SSA-LSSVM model

Abstract

Abstract Pile foundations in soft soil often encounter challenges regarding abnormal bearing capacity, significantly impacting the safety of engineering projects. The bearing capacity of pile foundations is influenced by various factors, often characterized by complexity and unpredictability. Therefore, this study proposes a comprehensive model, termed GWOASSA-LSSVM, based on grey wolf optimization (GWO)-enhanced sparrow search algorithm (SSA) and least squares support vector model (LSSVM). By introducing a grey wolf hierarchy to enhance the global search capability of the sparrow search algorithm and automating the optimization of parameters (γ, δ) in the LSSVM model using GWO, the GWOASSA-LSSVM predictive model is established. Evaluation metrics including correlation, and correlation between pile bearing capacity and the predictive target are considered, with pile bearing capacity as the predictive target. The GWOASSA-LSSVM model is compared with SSA-LSSVM, LSSVM, and back propagation neural network (BPNN). Results indicate that the GWOASSA-LSSVM model outperforms SSA-LSSVM, LSSVM, and BPNN across metrics such as coefficient of determination (R 2), variance accounted for (VAF), performance index (PI), index of agreement (IOA), index of scatter (IOS), mean absolute percentage error (MAPE), root mean square error (RMSE), and a20-index, demonstrating its capability for accurate prediction of pile bearing capacity.