Exploring the Potential of Machine Learning in Predicting Soil California Bearing Ratio Values

Abstract

Accurately predicting the California Bearing Ratio (CBR) of soil is vital for civil engineering projects as it determines soil strength and stability, crucial for designing safe and durable infrastructure. Conventional methods for calculating CBR values are both expensive and time-consuming, prompting the need for more efficient approaches. This study explores the use of advanced machine learning (ML) techniques to improve workflow and productivity in CBR prediction. Specifically, the Improved Arithmetic Optimization Algorithm (IAOA) and the Bonobo Optimization Algorithm (SBOA) are applied to enhance the Stochastic Gradient Boosting Regression (SGBR) model for predicting CBR values. The SGBR model, known for its ability to handle complex datasets and nonlinear interactions, is optimized to improve predictive accuracy. Performance metrics such as the coefficient of determination (R²), n10-index, and Root Mean Squared Error (RMSE) are used to assess the model's performance. After training, testing, and validation with relevant data, the optimized SGIA model (SGBR enhanced by IAOA) achieves impressive results, including an n10-index of 1.000, a root mean square error of 0.161, and an R² value of 0.981. These metrics demonstrate the SGIA model's capability to accurately forecast CBR values, offering a reliable, cost-effective alternative to traditional methods for soil evaluation in engineering applications.