- Research Article
- 10.1080/15397734.2025.2531077
Machine learning to estimate the California bearing ratio
- Jul 13, 2025
- Mechanics Based Design of Structures and Machines
- Xinyi Wang + 1 more +1
To assess the strength and stiffness of the subgrade, one of the most important parameters in pavement design is the California Bearing Ratio (CBR). Soil compaction and index properties influence CBR; therefore, further research was conducted to explore this relationship. CBR equation so that the thickness design of the flexible pavements can reach a certain probability. CBR estimation in this work is done by using Support Vector Regression (SVR). To enhance the predicted core model accuracy, Rune Kutta Optimization (PKO), Flying Foxes Optimization (FFO), and Henry Gas Solubility Optimization (HGSO) techniques have been employed as optimization methods. These sophisticated hybrid models were developed through an intentional effort to fuze these optimizers intentionally with the base models as a way to improve accuracy. When the SVR model is coupled with HGSO, the formed framework goes under the name of the SVFF framework; when it is combined with FFO, it goes under the name of the SVHG framework; and when connected with RKO, the framework takes the name of SVRK. As selected by the R 2 index, the best performance during the testing phase was given by the model SVRK, with a score of 0.963. The second-best and the best performance was that of the SVFF model, with an R 2 score of 0.957. During testing, the SVRK model outperformed the normal SVR model, which had a better RMSE value of 0.192, while the worst performance was by the latter model itself, with an RMSE value of 0.370.
Read more