A novel swarm intelligence optimized extreme learning machine for predicting soil shear strength: A case study at Hoa Vuong new urban project (Vietnam)

Abstract

In geotechnical engineering, soil shear strength is one of the most important parameters used in the design and construction of construction projects. However, determining this parameter in the laboratory is costly and time-consuming. Therefore, the main objective of this work is to develop a new alternative machine learning approach based on extreme learning machine (ELM) and Particle Swarm Optimization (PSO), namely PSO-ELM, for the shear strength prediction of soil for the Hoa Vuong new urban project in Nam Dinh province, North Vietnam. For this purpose, twelve soil parameters were collected on data from a survey of 155 soil samples to construct and validate the proposed model. We assessed the model's performance using the root-mean-square error (RMSE), the mean absolute error (MAE), and the coefficient of determination (R2). We compared the model's capability with five benchmark models, support vector regression (SVR), Gaussian process (GP), multi-layer perceptron neural network (MLP-NN), radial basis function neural network (RBF-NN), and the fast-decision tree (Fast-DT). The results revealed that the proposed PSO-ELM model yielded the highest prediction performance and outperformed the five benchmark models. It suggests that PSO-ELM can be an alternative method in estimating the shear strength of soil that would help geotechnical engineers reduce the cost of construction.